Article Search
닫기

Original Article

Split Viewer

International Journal of Fuzzy Logic and Intelligent Systems 2023; 23(4): 465-481

Published online December 25, 2023

https://doi.org/10.5391/IJFIS.2023.23.4.465

© The Korean Institute of Intelligent Systems

Data Analysis for Fuzzy Extreme Learning Machine

Archana P. Kale

Modern Education Society’s College of Engineering, Savitribai Phule Pune University, Maharashtra, India

Correspondence to :
Archana P. Kale (archana.mahantakale@gmail.com)

Received: August 17, 2022; Accepted: December 22, 2023

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/3.0/) which permits unrestricted noncommercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

Fuzzy extreme learning machine (F-ELM) one of the learning algorithm which is specifically used for uncertainty classification. Uncertainty classification is critical problem in the area of machine learning. In various real-time applications, the ambiguity is present in the input dataset itself which affects the systems generalization performance. Data (feature) analysis plays a major role in such types of problems. To solve the said problem in this paper, distance-based Relief algorithm and fuzzy extreme learning algorithm are used for data analysis and classification, respectively which contributes to Relief-based data analysis for F-ELM (RFELM++) and Relief-based data analysis for online sequential ELM (RFOSELM++) for batch mode and sequential input, respectively. Experimental results are calculated by using clinical dataset. Through the results, it is observed that RFELM++ produces increased accuracy in comparison with RELM++ for clinical dataset. The RFOSELM++ maintain accuracy by using 41.5% features as differentiate to OS-ELM for the UCI Repository dataset. The proposed RFOSELM++ algorithm is compared with similar already available sequential based algorithms. As a case study, a novel application of classification of nutrient deficiency and plant disease in which ambiguity presents is considered. Both proposed algorithms are exploited in this expert system which helps the remote farmer with expert advice. The proposed RFELM++ algorithm is tested and validated by using statistical methods.

Keywords: Uncertainty classification problem, Sequential problem, Internet of Things, Feature subset selection problem, Pattern classification

Classification, feature subset selection (FSS), and sequential problems are some of the critical problems in machine learning which has drawn substantial aspect from researchers in recent years. Due to the technological advances, data gets generated at an ever increasing space and the dimensionality or size of the datasets are increased by the day. Therefore, it is mandatory to design and develop the potent and adequate machine learning methods, that can be used to select, analyse and extract only the beneficial features. The existence of the dispensable and immaterial features in input may decreases the performance of systems. FSS is critical important pre-processing technique which selects only prominent (non-redundant and relevant) features. Pattern classification is a critical process which is utilize to classify the input dataset into one of the output classes.

As database and computer technologies advance speedily, various advancing challenges appear with respect to FSS and classification problems as follows:

  • • Improvement in accuracy by using the same number of attributes;

  • • Maintain accuracy by downsizing the attributes;

  • • Improvement in accuracy by reducing the number of attributes.

FSS problem has drawn considerable attention from researchers in current years due to the existence of a huge attributes. Many real-time applications contain high-dimensional dataset. These thousands of features are usually provided to the learning algorithms for the classification task.

In many real-time applications, the attributes present in the dataset are ambiguous. Therefore, it becomes very challenging to classify the instance of input. Such kind of the problems in which the ambiguity presents in the attributes of the input dataset [1] are come under the uncertainty classification problem (UCP). Uncertainty classification is used for one class classification [2], binary classification problems [3,4] and imbalance classification problem via support vector machine (SVM) [5]. A gradient approach is developed in Naive Bayes (NB) for value uncertainty classification learning by Lee [6]. Zhang and Ji [1] developed fuzzy extreme learning machine (F-ELM) for uncertainty classification problem.

F-ELM is the latest classifier which associates the working of fuzzy inference system (FIS) and ELM. For uncertainty classification problem F-ELM algorithm is mostly used [1, 7]. F-ELM uses a membership function for required input which is represented by linguistic variables.

Many research papers show the different FSS techniques for various classification problems. However, FSS with incremental learning for uncertainty classification problem is absent in the vast literature survey.

The original F-ELM and ELM are the batch learning mode which considers that the data is available initially. So, F-ELM and ELM classifiers are unable to handle sequent or serial input. So that, online sequential extreme learning machine (OS-ELM) algorithm is presented by Nan-Ying et. al. for sequent input. Three layers are available for ELM and OS-ELM: input, hidden, and output. Randomly initialization of weights is carried out from the first (input) to second (hidden) layer and logically modulated the weights from the second (hidden) to third (output) layer. On account of random initialization, generalized performance of system may degrade. So that it is necessary to select the prominent features.

To distinguish said problem, in this paper incremental F-ELM (RFELM++) and incremental fuzzy online sequential ELM (RFOSELM++) for the UCP are proposed for batch mode and sequential input respectively. Thus, the key intent of this paper is to design RFELM++ and RFOSELM++ algorithm by using the activation function like radial basis function (RBF) and additive function which is missing in the vast literature survey. Another aspect of the paper is to design an IoT based expert system - Precision Agriculture by exploiting the proposed algorithms.

The overall structure of the paper is the following: in Section 2, the literature survey is detailed in the related work. The proposed incremental F-ELM and RFOSELM++ algorithms are depicted in Section 3. The experimental observations and results of clinical datasets are discussed and compared in Section 4. The validation and testing part of the proposed algorithm are discussed in detail in Section 5 as the results and discussions. Section 6 outlined the conclusion with further research direction.

2.1 Feature Subset Selection

In various growing areas FSS plays a very critical task. The key role of FSS is to choose prominent/optimal (relevant and non-redundant) feature subset which produces similar or improved (in some of the cases) results as compared to the prime or base data set in which all features are present. It is very hard to decide the significance of attributes without preparatory information [8]. Therefore, FSS is an prime step to select prominent features for efficient machine learning applications [9, 10]. The primary intention of using any FSS techniques is for [11]:

  • • Improvement in classification accuracy,

  • • Improvement in prediction performance,

  • • Improvement in model interpretability,

  • • Dimensionality reduction,

  • • Storage management.

2.1.1 FSS techniques

The few of the FSS techniques which are mostly used are: filter, wrapper, and hybrid.

Filter techniques rank the features or select feature subset independently of the learning algorithm. Filter techniques are easy to interpret and fast. The characteristics of filter techniques are considered independently, totally independent of which learning algorithm is used etc. For class separability, the evaluation criteria is required which is a classical criterion of the filtering present in the literature. The criteria like distance, information and reliant are mostly utilize to classify the features.

Wrapper techniques use a learning algorithm for FSS, training is required for every feature subset. So, wrapper techniques are computationally heavy and dependent on the type of classifier used. The characteristics of wrapper techniques are computationally expensive, use only heuristic search and dependent on the learning algorithm.

Hybrid techniques use both filter and wrapper techniques. Ranking-wrapper techniques [12], Relief-F with sequential backward search [13] and boosting-based hybrid for feature selection (BBHFS) [14] are few of hybrid techniques.

Tahir and Loo [15] used Relief-F method for feature selection in continual learning framework. Various optimization algorithms like particle swarm intelligence [16] is used to solve feature selection problem and multi-objective particle swarm optimization algorithm [17] is used to solve bioprocess application problems and tumor treatment problems.

2.2 Pattern Classification

Pattern classification is specially used for decision making task. It classifies input entity to one of the predetermined target outputs. A supervised classification is specially used to categorised the given input data into one of the present target classes [18]. Having the data D = xi, yi, where i = 1, …, K and xiRn is a single pattern or instance with n features, yi ∈ {1, 2, …, m}. yi is the total number of target class which is represented by m and K represents total number of patterns (D); a problem of pattern classification is the determination of a mapping model M(.), where M(xi) = yi [19]. Neural network (NN) is a robust computational tool which is specifically used for classification. Various NN based classifiers–backpropagation NN (BPNN), multilayer perceptron (MLP), radial basis function (RBF), decision tree, support vector machine (SVM), C4.5, fuzzy NN (FNN), convolutional neural network (CNN), ELM, etc.–are required for classification. Tao et al. [20] designed a deep CNN for detecting the defects in insulators. Wen et al. [21] developed a deep transfer learning (DTL) method for fault diagnosis. ELM is used in breast cancer detection with the help of multilayer fuzzy expert system which was developed by Mojrian et al. [22]. Basically, the classifier plays a very crucial act in system overall performance. The CNN classifier faces various problems like over fitting problem, local maxima and minima problem, etc. Therefore, classifier selection is a challenging task.

2.2.1 ELM

The original ELM classifier is basically designed for a batch mode. ELM is acknowledged that the complete input data is available previously. However, many real applications require to handle the data sequentially. Therefore, Nan-Ying Liang et. al. have designed OS-ELM algorithm for handling the sequent or serial input. OS-ELM randomly assigns all the weight values in-between first (input) and second (hidden) layer and the weight values in between second (hidden) to third (output layer) are analytically tuned [2325] as demonstrated in Figure 1 [26]. The global generalization performance may degrade due to the existence of non-prominent attributes and random initialization. Therefore, it is necessary to choose a prominent feature subset for sequential input. However, ELM is insufficient to provide a solution for the sequential and UCP.

2.2.2 OS-ELM

OS-ELM is exploited in various applications like estimation of hematocrit [27], DWT domain watermarking [28], traffic profiling [29], etc. However, the literature survey lacks the exploitation of incremental learning ability by using prominent or optimal feature subset selection for sequent or serial input.

2.2.3 F-ELM

Traditional ELM is unable to solve some real-time time problems like the UCP and imbalance problem [1]. So, F-ELM is designed by Zhang and Ji [1] to solve specially UCP. However, in order to solve the UCP for sequential input OS-ELM is designed [30].

Ontiveros-Robles et al. [31] used type II fuzzy classification for non-singleton and also designed hybrid classification by using SVM [32]. Rubio et al. [33] used fuzzy clustering for unsuprvised method.

2.3 Incremental Learning

Incremental learning is used in various ways like incrementally add hidden nodes [34, 35], incrementally add streaming data [36], incremental FSS [3740], etc. The exploitation of incremental learning in terms of incremental FSS for sequential input is also missing in the literature survey.

Incremental learning comes in various forms and definitions as

  • • Incrementally add streaming data: Incremental learning is a learning for the data which appear sequentially or over time [36, 41]. Alexander Gepperth describes online learning, incremental learning and concept drift for the supervised learning paradigm in detail.

  • • Incrementally add hidden nodes: Incremental learning defines as an incrementally insertion of the new neurons to the hidden layer. Error-minimized ELM (EM-ELM) [42], Convex incremental ELM (CI-ELM) [35] and uncremental extreme learning machine (RELM++) [34] are some of the examples of incremental learning. Yang et al. [43] designed a bidirectional extreme learning machines (B-ELM) algorithm for regression.

  • • Incremental feature subset selection: Incremental learning in the form of incremental feature subset selection is used in many papers like [3740]. Baluja [44] have developed a population-based incremental learning method which integrated learning and the random search based function optimization [44]. Tahir and Loo [15] used an incremental ELM is used for food recognition. Wang et al. [45] have designed incremental wrapper based gene selection with Markov blanket. Incremental wrapper subset selection (IWSS) uses a filter measure for feature ranking by using a sequential search [46]. Then, a SFS is used for incrementally insertion of the features.

Wang et al. [47] designed a framework for smart contracts based on a novel six-layer architecture.

This section is categorized into two subsections, RFELM++ and RFOSELM++ algorithms.

3.1 RFELM++ Algorithm

In this section, incremental learning is used to resolve the uncertainty classification problem. As already discussed in Section 2, F-ELM specifically uses for the uncertainty classification problem. Hence, RFELM++ algorithm is initiated. The paradigm of RFELM++ is broadly categorised into four subsystems: input, incremental learning for FSS, fuzzification, and classification.

Input subsystem

In this section, the detailed description of the input is described in detail. In order to resolve the UCP, it is necessary to use the datasets in which uncertainty is present. Therefore, Statlog Heart Disease (SHD) and Pima Indian Diabetes (PID) datasets are used [48]. The PID data set contains total 768 instances and 8 attributes. The datasets consists of several medical predictor variables and one target variable, outcome. Predictor variables includes the number of pregnancies the patient has had, their body mass index (BMI), insulin level, age, and so on. The SHD data set contains 270 instances and 13 attributes. Additionaly, the Cleveland Heart Disease (CHD) dataset is used for multiclass classification.

Incremental learning for FSS

The generalization performance of any system is depend on the input. If the input to the system is right the performance of system gets increased. Therefore, it is very important task to provide a right or correct input features to the system.

Prominent feature analysis

In data preprocessing FSS algorithm is used. To select only prominent features relief algorithm is used. Relief algorithm is one of the FSS algorithms based on distance measure. The measure of distance (D) is basically used to search the features which can differentiated between two target classes. If the D is equal to 0, then the two attributes are identical. The Relief algorithm is used the joint relationship to compute weights with the output [49, 50] which is famous due to it’s simplicity and effectiveness. In Relief (feature weight-based algorithm), a statistical method is used to choose the relevant attributes [51].

After applying relief algorithm on PID and SHD datasets, we get feature order. For example, PID dataset has 8 attributes like age, DPF, BMI, insulin, skin, BP, Glu, and pregnancy. After Relief algorithm we get ⟨2, 6, 4, 8, 1, 7, 5, 3⟩ feature order. With this order it is understood that 2 is the most required or more relevant feature whereas 3 is the least required feature.

As per already described in the previous section, prominent feature order is calculated by considering distance measures i.e., Relief algorithm. Suppose the attribute order for SHD and PID data sets is ⟨13, 12, 3, 9, 10, 11, 2, 1, 7, 4, 5, 6, 8⟩ and ⟨2, 6, 8, 1, 7, 5, 4, 3⟩ respectively. The organised attributes are incrementally inserted one by one by using SFS for sequential space. Experimental observation shows that from all the subsets {1, 2, 3, 4, 5, 7, 9, 10, 11, 12, 13} and {1, 2, 4, 5, 6, 7, 8} are the prominent feature subset for SHD and PID datasets, respectively, as it provides improved classification accuracy.

Fuzzification and classification

The attributes of PID and SHD datasets are translated into linguistic attributes (fuzzified features) with the help of trapezoidal membership function [52]. The PID (8 attributes) and SHD (13 attributes) datasets are translated into 25 and 39 linguistic attributes, respectively. Input dataset is distributed into two datasets like training with 70% and testing with 30 %. Hence for PID ans SHD datasets 577 and 177 instances are used for traing and 577 and 177 instances are used for testing, respectively. By using trapazoidal membership function the features are converted into fuzzified features. For classification ELM is used.

3.2 RFOSELM++ Algorithm

ELM is a batch learning mode classifier which is not suitable for Sequential input. In this section, incremental learning for FSS is used to solve the UCP as well as sequential problem. Hence, incremental FSS techniques for the UCP is proposed for sequential input which contributes to RFOSELM++ algorithm.

In RFOSELM++, the ranked features are inserted sequentially i.e., one by one. The accuracy is calculated for all feature subsets which are equal to the number of attributes. The proposed RFOSELM++ learning algorithm is applied for every subset and predicted accuracy is calculated. The paradigm of RFOSELM++ is mostly same as RFELM++. Only the classification subsystem is different one. Through observations, the prominent feature subset is one which provides maximum accuracy.

For experimental observation, Waikato Environment for Knowledge Analysis (WEKA) and MATLAB R2014a are used. Precision, Recall and classification accuracy are used as an evaluation measure. The calculations of evaluation measures are depend upon the contingency table or confusion matrix of classifier. Confusion matrix contains both true and false positives (TP/FP) also true and false negatives (TN/FN) [52]. FP correspond to negative samples falsely identified as positive. FN correspond to positive samples falsely resembles as negative. TP are the samples perfectly identified as positive. TN refer to negative samples perfectly identified as negative. The perfectness of identified examples are measured by using precision. The measure of the all positive identified examples is called recall [53]. Accuracy, precision and recall (sensitivity) are measured by evaluating the respective values in Eqs. (1) to (3), respectively

Acc=TP+TNTP+FP+TN+FN,Pre=TPTP+FP,Rec=TPTP+FN.

In this section, the incremental learning capability for batch and sequential mode is exploited i.e., incremental learning for F-ELM (RFELM++) algorithm and incremental learning for fuzzy online sequential extreme learning machine, respectively.

4.1 RFELM++ (Prominent Feature Subset)

The experimental results of the proposed RFELM++ algorithm are enumerated. In order to prove the effectiveness of an initiated algorithm the same steps are implemented for ELM classifier (RELM++). The results of both RELM++ and RFELM++ are compared with sigmoidal activity function by considering PID and SHD datasets. The overall comparison of RELM++ and RFELM++ for PID dataset is as shown in Figure 2. The same results are evaluated for SHD dataset also. Through the results, it is observed that RFELM++ produces 6.266% and 0.855% increased accuracy in comparison of RELM++ for PID and SHD datasets consequently.

The overall performance of the FELM classifier is analysed by considering the precision, accuracy and recall. Figure 3 shows an overall comparison of performance for ELM and F-ELM with all features. The results are calculated by using all activity functions like sigmoid function (sig), radial basis function (rbf), sine function (sin), hardlim function (hardlim), triangular basis function (tribas) for SHD and PID datasets. By observations it is apprise that, F-ELM issues an improved performance as compared to ELM classifier.

Various FSS methods like half selection (HS), mean selection (MS), particle swarm optimization (PSO) based FS, neural network for threshold, self-regulated learning (SR-PSO) [54] are used to select prominent features. Table 1 indicates the output of all given FSS methods with ELM and FELM classifiers. For evaluation, the sigmoidal activation function is used for PID dataset. From experimented results, it is noticed that the F-ELM achieves 4.65% improved classification accuracy in comparison with ELM for the same number of feature. The result for RFELM++ algorithm for multiclass classification problem for CHD dataset is as shown in Figure 5.

In order to prove the effectiveness of the proposed RFELM++ algorithm, experiment results are derived and analysed by using the traditional classifiers–MLP, random forest (RF), Bayes net (BN), SVM, NB, RBF, and J4.8–by using prominent features. The prominent features are selected by using Relief algorithm. Five-fold cross-validation method is used for evaluation. Figure 4 shows the variation of the proposed RFELM++ algorithm with existing classifiers by using relief FSS algorithm for PID and SHD data set. It is evidently prove that the proposed RFELM++ algorithm enables maximum predictive accuracy with minimum number of features.

4.2 RFOSELM++ (Prominent Feature Subset)

PID dataset is used as an input dataset. Normalization is used in pre-process. The algorithm uses relief algorithm (distance measure) for ranking the features. IWSS [39, 40] is used which acquires a ranking of the features over to the target class. IWSS is used SFS search over the ranked attributes by incrementally inserting those features one by one. The wrapper way is used to measure the relevance of the new attribute. The classification accuracy is calculated by using OS-ELM classifier.

Experimental results of incremental attribute learning for sequential input is calculated for both ELM and F-ELM classifier. Table 2 shows the parameters required for the evaluation. Figures 6 and 7 show the comparison of RELM++ and RFELM++ for training and testing accuracy, respectively. Through observation, it is noticed that the proposed RFELM++ issues an improved as well as stable (constant) results as compared to RELM++.

In order to prove the validation of the proposed RFELM++ algorithm, experimental observations need to be validated and hypothetically tested [55]. F-ELM classifier is used after the feature selection by using statistical methods. The PID and SHD datasets are categorized into four samples like S4 (SHD ELM), S3 (PID ELM), S2 (SHD F-ELM) and S1 (PID F-ELM). The comparative analysis of before and after relief feature selection algorithm is as shown in Figure 8. Through the results it is noticed that, RFELM++ algorithm with feature subset produces improved performance as compared to with all features.

The correlation analysis technique is used to compute the robustness of the relationship between 2 or more variables. All the time the values of correlation are in between −1 to +1. The abbreviation of +1 is that the two variables are strongly related with positive linear, −1 is that two variables are resolutely related with negative linear and 0 is that no relationship in between the two variables. According to Evans classification, the correlation coefficient are broadly categorized into very strong, strong, moderate, weak, very weak [56].

The c (Pearson’s correlation coefficient) is given by

cxy=pq[n.p2-(p)2].[n.q2-(q)2],

where ‘p’ is metric before feature selection and ‘q’ is metric after feature selection.

The coefficient (c = 0.6177) is calculated by using all features and Relief FS. Figure 8 illustrates the computation of evaluation metrics accuracy for ELM and F-ELM learning algorithms. Through this computation, it proves that the relationship in between the F-ELM (all features) and RFELM++ (feature subset) is strong as per Evans classification [56].

When the relationship (c) is strong, then the Student’s paired t-test is utilised for comparison in-between two discrete techniques. Paired t-test (pt-value) is calculated by using Eq. (5). To check the competence of the initiated RFELM++ algorithm, Student’s paired t-test is used. Based on the value of r = 0.6177, the t-value is calculated, i.e., −1.097.

pt=qq2-(q)2/n-1,

where ‘q’ is the difference between ‘a’ (before applying FSS algorithm) and ‘b’ (after applying FSS algorithm) (qi = aibi).

Hypothesis of testing [57] RFELM++:

  • H0: No variance between the generalization performance of F-ELM (with all features or before feature selection) and RFELM++ (after feature selection)

  • H1: Variance between the generalization performance of F-ELM (with all features or before feature selection) and RFELM++ (after feature selection)

The pt value i.e., Student’s t-distribution for the degree of freedom 4 is (2.776, ∝= 0.05) and (4.604, ∝= 0.01) [58]. By observation, it is notified that there is no variance between the generalization performance of F-ELM (with all features or before feature selection) and RFELM++ (after feature selection) for four samples, where ∝ is the level of significance. The main reasons behind the improved performances of the proposed algorithm are the removal of irrelevant and redundant features from the input dataset before weighted classification.

5.1 Benchmark Comparison

For performance comparison, accuracy of RFELM++ by using prominent feature subset is compared with F-ELM [59], modified fuzzy min-max neural network-FIS (MFMM-FIS) [60] and enhanced generalized adaptive resonance theory FIS (EGART-FIS) [61] for the PID dataset. Table 3 describes the comparative classification accuracy for the same. With the comparative analysis, it is observed that the testing and training accuracy is 83.47% to 86.80%, respectively by using only 62.5% features.

5.1.1 RFELM++ and the existing classifiers with different FSS methods

With the intention to evince the effectiveness of the proposed RFELM++ algorithm, it is mandatory to have comparison the result of RFELM++ with the similar approach like NN for threshold selection (NN) [62,63], multilayer NN (MLNN) by using Levenberg–Marquardt (LM) [64], generalized discriminant analysis least square (GDA-LS) [65], HS method, least square-ELM (LS-ELM) [65], linear discriminant analysis and Morlet wavelet (LDA-MW), probabilistic neural network (PNN) [64], MS methods. Table 4 shows RFELM++ and other classifiers comparison for PID input dataset. Table 5 illustrates the same comparison for SHD dataset with evolutionary product-unit NN (EPUNN), MS, evolutionary sigmoidal unit NN (ESUNN), HS, multilogistic regression by means of evolutionary product-unit NN (MR+EPUNN) for SHD dataset. Through comparative average analysis, it is noticed that the introduced RFELM++ issues 8.82% and 6.63% increased accuracy by reducing 6.41% and 10.71% for SHD adn PID datasets, respectively.

The comparative analysis in between F-ELM and ELM by using different FSS algorithms is as shown in Figure 9 for PID dataset. HS, PSO, MS, NN, T-test and self-regulated learning (SR-PSO) [54] are used as FSS algorithms. For PID dataset, the comparison of ELM and F-ELM classifier in terms of accuracy measure by using the feature subset as shown in Figure 9 in which F-ELM produces increased performance for the same number of features.

The introduced RFELM++ algorithm is exploited in the real time application which is represented as a case study in the following section.

Agriculture is the backbone of Indian economy where approximate 60% of people are dependent directly or indirectly on agriculture. The expert advice is required for distinguishing the plant disease damage and nutrient imbalance. It is observed that, the conventional judgmental analysis is not enough while deciding the quantity of chemical or fertilizer to be used. The mis-proportional dose harms directly the health of the crop and hence the living beings. To overcome the said problem, this section is developed an IoT-based precision agriculture expert system by inferring proposed algorithms with improving the accuracy by reducing the features or by selecting the prominent features. Grape leaf diseases like downy mildew and nutrient deficiency which contains ambiguity is considered as a case study. Both are spread with yellow spot on leaf. The conceptual framework for the proposed IoT-based precision agriculture expert system is as shown in Figure 10.

In IoT-based precision agriculture expert system, the system is partitioned into three rough phases namely image processing, FSS by using incremental learning and classification. In image processing, the acquisition step is used to input the image. The image is acquired by using cell phone or any digital camera. The resolution of the image is 96 dpi (dots per inch). For interfacing, the ARDUINO hardware is used with various sensors like temperature, humidity etc. The leaf image is used to extract the feature points and are stored in the database.

Data normalization is used as the preprocessing method. In image analysis, image segmentation and feature extraction methods are considered. For image segmentation, region-based segmentation is used to separate the healthy and diseased region by using leaf color. The features extracted from an images are color, shape and texture [71]. Once features are extracted, the database is generated with integrated information from temperature and humidity sensors. On that created dataset the proposed algorithms are applied for feature selection and classification.

In order to increase the generalization performance of the system, the proposed RELM++ and RFOSELM++ algorithms are exploited. The performance of precision agriculture is evaluated and tested by using RELM++ and RFOSELM++ algorithms. The comparison in-between RELM++ and ELM for proposed system (plant disease and nutrient deficiency) is as shown in Figure 11 in terms of testing accuracy. Through this experimentation, it is intended that the RELM++ produces 5.71% increased accuracy and RFOSELM++ maintains (same accuracy) the accuracy with only 27.27% features. However, for Tribas activation function the classical method provides good results as compared to the proposed one. Tribas provided an optimal result only for certain number of hidden neurons for ELM [72].

The prime intention of this paper is to exploit the two proposed algorithms like RFELM++ and RFOSELM++ into the new developed IoT-based precision agriculture Expert system. The expert system helps farmers in decision making. The experimental results for ELM, OS-ELM, RELM++, RFOSELM++, RFELM++ and RFOSELM++ are calculated. RFELM++ produces 6.266% and 0.855% increased accuracy in comparison of RELM++ for PID and SHD datasets consequently. Additionally, various FSS methods like like HS, MS, PSO based FS, neural network for threshold, self-regulated learning (SR-PSO) [54] are used to select prominent features. Through the experimented results, it is noticed that the F-ELM achieves 4.65% improved classification accuracy in comparison with ELM for the same number of feature. The proposed RFOSELM++ maintain the accuracy by using 41.5% features as differentiate to OS-ELM for UCI Repository dataset. For benchmark problem, the results of initiated RFELM++ algorithm is compared with the similar existing approaches. From the comparative average analysis, it is noticed that the introduced RFELM++ issues 8.82% and 6.63% increased accuracy by reducing 6.41% and 10.71% for SHD and PID datasets, respectively. Table 3 describes the comparative classification accuracy for the same. With the comparative analysis, it is observed that the testing and training accuracy is 83.47% to 86.80%, respectively by using only 62.5% features. Hypothetically it is proved that there is no variance between the generalization performance of F-ELM (with all features or before feature selection) and RFELM++ (after feature selection) for four samples. A novel application, IoT-based precision agriculture expert system, is developed by exploiting the proposed algorithms RFELM++ and RFOSELM++. Currently, the algorithms are used for binary classification in future it can be used for multi-class classification. The precision system can be improved by considering the rain removal [73] and by using CNN [74].

Fig. 1.

Extreme learning machine.


Fig. 2.

RELM++ and RFELM++ by using relief algorithm.


Fig. 3.

Comparison of ELM and F-ELM: (a) accuracy of PID dataset, (b) accuracy of SHD dataset, (c) precision of PID dataset, (d) precision of SHD dataset, (e) recall of PID dataset, and (f) recall of SHD dataset.


Fig. 4.

Accuracy measure comparison of F-ELM with various classifiers by using Relief-FSS algorithm: (a) input PID dataset and (b) input SHD dataset.


Fig. 5.

Evaluation performance: (a) CHD accuracy, (b) CHD precision, and (c) CHD recall.


Fig. 6.

ELM and F-ELM performance comparison with incremental learning for training.


Fig. 7.

ELM and F-ELM performance comparison with incremental learning for testing.


Fig. 8.

Performance metrics of RFELM++ before and after using Relief-FSS algorithm.


Fig. 9.

F-ELM and ELM in combination with different FSS methods for PID dataset.


Fig. 10.

Conceptual framework for the IoT-based precision agriculture expert system.


Fig. 11.

ELM and RELM++ (testing accuracy) for plant disease and nutrient deficiency.


Table. 1.

Table 1. Comparative results of ELM and F-ELM with different FS methods for input PID dataset.

FSS MethodsELMF-ELMFeature subset
MS78.2679.132, 6, 8
HS; l76.0882.61, 2, 6, 8
NN for threshold selection78.2682.61, 2, 6, 8
PSO78.2682.61, 2, 6, 8
SRLPSO78.2679.132, 6, 8
t-test76.5283.471, 2, 6, 7, 8
Without FSS Method71.7380.43All
Avg76.7681.42-

Table. 2.

Table 2. Parameters for RFELM++.

ParameterFeatures
Input neuronSame as input features
Output neuron2
Hidden neuron10:5:200
Activity functionSigmoidal
Sequential mode1 by 1, 20 by 20, 10 by 30
Dataset divisionTraining (70%)-Testing (30%)

Table. 3.

Table 3. Comparison of RFELM++ with existing learning algorithms for input PID dataset.

Methods/algorithmAccuracy of trainingAccuracy of testingTotal number of attributes used
RFELM++86.8083.4705 (62.5%)
F-ELM [59]75.3574.0908 (100%)
MFMM-FIS [60]n/a72.9208 (100%)
EGART-FIS [61]n/a73.0508 (100%)
MFMM-FIS [62]n/a72.9208 (100%)

Table. 4.

Table 4. RFELM++ and existing learning algorithms comparison for input PID dataset.

Method/algorithmAccuracy (testing)Total number of attributes usedFeature subset
RFELM++83.4705{1,2,6,7,8}
RELM++76.5205{1,2,6,7,8}
EGART-FIS [61]73.0508{1,2,3,4,5,6,7,8}
F-ELM [59]74.0908{1,2,3,4,5,6,7,8}
MFMM-FIS [60]72.9208{1,2,3,4,5,6,7,8}
LS-ELM [65]78.2104{1,2,6,8}
GDA-LS-ELM [65]79.1605{1,2,6,7,8}
MLNN with LM [64]79.6204{1,2,6,8}
PNN [64]78.0503{2,6,8}
MS [62]76.0403{2,6,8}
HS [62]75.9104{1,2,6,8}
NN [62]76.0403{2,6,8}
GP-KNN [66]80.5008{1,2,3,4,5,6,7,8}
Gini-Fuzzy [67]75.8008{1,2,3,4,5,6,7,8}
FMNN-CART-RF [25]78.3908{1,2,3,4,5,6,7,8}
Generic statistical approach [68]78.0408{1,2,3,4,5,6,7,8}
Average76.8426.79-

Table. 5.

Table 5. RFELM++ and existing learning algorithms comparison for Input SHD dataset.

Method/algorithmAccuracy (testing)Number of featuresFeature subset
RFELM++92.5906{3,9,10,11,12,13}
RELM++90.7405{3,9,10,12,13}
ESUNN [69]83.2205{3,8,9,11,12}
EPUNN [69]81.8904{8,9,11,12}
MR+EPUNN [70]83.1205{8,9,11,12,13}
MS [62]84.4406{3,8,9,11,12,13}
HS [62]84.8107{3,8,9,10,11,12,13}
NN [62]85.1904{3,11,12,13}
Average83.7760.26-

Table. 6.

Table 6. Evaluation comparative analysis for ELM and F-ELM with different FSS for input PID dataset.

FSS MethodELMF-ELMFeature subset
LS [65]78.2182.6{1,2,6,8}
GDA-LS [65]79.1682.17{1,2,6,7,8}
MLNN with LM [64]79.6282.6{1,2,6,8}
PNN [64]78.0579.13{2,6,8}
LDA-MW [62]89.7482.17{1,2,6,7,8}
MS [62]76.0479.13{2,6,8}
HS [62]75.9182.6{1,2,6,8}
NN [62]76.0479.13{2,6,8}

  1. Zhang, WB, and Ji, HB (2013). Fuzzy extreme learning machine for classification. Electronics Letters. 49, 448-450. https://doi.org/10.1049/el.2012.3642
    CrossRef
  2. Krawczyk, B, Wozniak, M, and Herrera, F . Weighted one-class classification for different types of minority class examples in imbalanced data., Proceedings of IEEE Symposium on Computational Intelligence and Data Mining (CIDM), 2014, Orlando, FL, USA, Array, pp.337-344. https://doi.org/10.1109/CIDM.2014.7008687
    CrossRef
  3. Juan, J, and Cubero, C . Weighted classification using decision trees for binary classification problems., Proceedings of the Spanish Computer Science Congress (CEDI 2007, II Congreso Español de Informática), 2007, Zaragoza, Spain.
  4. Eitrich, T, and Lang, B . On the advantages of weighted l1-norm support vector learning for unbalanced binary classification problems., Proceedings of 2006 3rd International IEEE Conference on Intelligent Systems, 2006, London, UK, Array, pp.575-580. https://doi.org/10.1109/IS.2006.348483
    CrossRef
  5. Hwang, JP, Park, S, and Kim, E (2011). A new weighted approach to imbalanced data classification problem via support vector machine with quadratic cost function. Expert Systems with Applications. 38, 8580-8585. https://doi.org/10.1016/j.eswa.2011.01.061
    CrossRef
  6. Lee, CH (2015). A gradient approach for value weighted classification learning in naive bayes. Knowledge-Based Systems. 85, 71-79. https://doi.org/10.1016/j.knosys.2015.04.020
    CrossRef
  7. Kale, AP, and Sonavane, S (2018). PF-FELM: a robust PCA feature selection for fuzzy extreme learning machine. IEEE Journal of Selected Topics in Signal Processing. 12, 1303-1312. https://doi.org/10.1109/JSTSP.2018.2873988
    CrossRef
  8. Yu, L, and Liu, H (2004). Efficient feature selection via analysis of relevance and redundancy. Journal of Machine Learning Research. 5, 1205-1224.
  9. Dash, M, and Liu, H (1997). Feature selection for classification. Intelligent Data Analysis. 1, 131-156. https://doi.org/10.1016/S1088-467X(97)00008-5
    CrossRef
  10. Forman, G (2003). An extensive empirical study of feature selection metrics for text classification. Journal of Machine Learning Research. 3, 1289-1305.
  11. Bharathi, P, and Subashini, P . Optimal feature subset selection using differential evolution with sequential extreme learning machine for river ice images., Proceedings of IEEE Region 10 Conference (TENCON), 2015, Macao, China, Array, pp.1-6. https://doi.org/10.1109/TENCON.2015.7372861
    CrossRef
  12. de Assis Boldt, F, Rauber, TW, Varejao, FM, and Ribeiro, MP . Fast feature selection using hybrid ranking and wrapper approach for automatic fault diagnosis of motorpumps based on vibration signals., Proceedings of 2015 IEEE 13th International Conference on Industrial Informatics (INDIN), 2015, Cambridge, UK, Array, pp.127-132. https://doi.org/10.1109/INDIN.2015.7281722
    CrossRef
  13. Tiwari, S, and Singh, B . A hybrid approach for feature selection., Proceedings of 2015 3rd International Conference on Image Information Processing (ICIIP), 2015, Waknaghat, India, Array, pp.277-280. https://doi.org/10.1109/ICIIP.2015.7414780
    CrossRef
  14. Das, S . Filters, wrappers and a boosting-based hybrid for feature selection., Proceedings of the Eighteenth International Conference on Machine Learning (ICML), 2001, Williamstown, MA, USA, pp.74-81.
  15. Tahir, GA, and Loo, CK (2020). An open-ended continual learning for food recognition using class incremental extreme learning machines. IEEE Access. 8, 82328-82346. https://doi.org/10.1109/ACCESS.2020.2991810
    CrossRef
  16. Zhang, T, Ding, B, Zhao, X, and Yue, Q (2018). A fast feature selection algorithm based on swarm intelligence in acoustic defect detection. IEEE Access. 6, 28848-28858. https://doi.org/10.1109/ACCESS.2018.2833164
    CrossRef
  17. bin Mohd Zain, MZ, Kanesan, J, Chuah, JH, Dhanapal, S, and Kendall, G (2018). A multi-objective particle swarm optimization algorithm based on dynamic boundary search for constrained optimization. Applied Soft Computing. 70, 680-700. https://doi.org/10.1016/j.asoc.2018.06.022
    CrossRef
  18. Zhang, GP (2000). Neural networks for classification: a survey. IEEE Transactions on Systems, Man, and Cybernetics Part C (Applications and Reviews). 30, 451-462. https://doi.org/10.1109/5326.897072
    CrossRef
  19. Tsipouras, MG, Voglis, C, and Fotiadis, DI (2007). A framework for fuzzy expert system creation: application to cardiovascular diseases. IEEE Transactions on Biomedical Engineering. 54, 2089-2105. https://doi.org/10.1109/TBME.2007.893500
    Pubmed CrossRef
  20. Tao, X, Zhang, D, Wang, Z, Liu, X, Zhang, H, and Xu, D (2020). Detection of power line insulator defects using aerial images analyzed with convolutional neural networks. IEEE Transactions on Systems, Man, and Cybernetics: Systems. 50, 1486-1498. https://doi.org/10.1109/TSMC.2018.2871750
    CrossRef
  21. Wen, L, Gao, L, and Li, X (2019). A new deep transfer learning based on sparse auto-encoder for fault diagnosis. IEEE Transactions on Systems, Man, and Cybernetics: Systems. 49, 136-144. https://doi.org/10.1109/TSMC.2017.2754287
    CrossRef
  22. Mojrian, S, Pinter, G, Joloudari, JH, Felde, I, Szabo-Gali, A, Nadai, L, and Mosavi, A . Hybrid machine learning model of extreme learning machine radial basis function for breast cancer detection and diagnosis; a multilayer fuzzy expert system., Proceedings of 2020 RIVF International Conference on Computing and Communication Technologies (RIVF), 2020, Ho Chi Minh City, Vietnam, Array, pp.1-7. https://doi.org/10.1109/RIVF48685.2020.9140744
    CrossRef
  23. Zhang, Y, Wu, J, Cai, Z, Zhang, P, and Chen, L (2016). Memetic extreme learning machine. Pattern Recognition. 58, 135-148. https://doi.org/10.1016/j.patcog.2016.04.003
    CrossRef
  24. Huang, GB, Zhu, QY, and Siew, CK (2006). Extreme learning machine: theory and applications. Neurocomputing. 70, 489-501. https://doi.org/10.1016/j.neucom.2005.12.126
    CrossRef
  25. Seera, M, and Lim, CP (2014). A hybrid intelligent system for medical data classification extreme learning machine: theory and applications. Expert System Application. 41, 2239-2249. https://doi.org/10.1016/j.eswa.2013.09.022
    CrossRef
  26. Karakaya, G, Galelli, S, Ahipasaoglu, SD, and Taormina, R (2016). Identifying (quasi) equally informative subsets in feature selection problems for classification: a max-relevance min-redundancy approach. IEEE Transactions on Cybernetics. 46, 1424-1437. https://doi.org/10.1109/TCYB.2015.2444435
    Pubmed CrossRef
  27. Huynh, HT, Won, Y, and Kim, J (2015). Hematocrit estimation using online sequential extreme learning machine. Biomedical Materials and Engineering. 26, S2025-S2032. https://doi.org/10.3233/BME-151507
    Pubmed CrossRef
  28. Singh, RP, Dabas, N, Chaudhary, V, and Nagendra, (2016). Online sequential extreme learning machine for watermarking in DWT domain. Neurocomputing. 174, 238-249. https://doi.org/10.1016/j.neucom.2015.03.115
    CrossRef
  29. Singh, R, Kumar, H, and Singla, R (2015). An intrusion detection system using network traffic profiling and online sequential extreme learning machine. Expert Systems with Applications. 42, 8609-8624. https://doi.org/10.1016/j.eswa.2015.07.015
    CrossRef
  30. Rong, HJ, Huang, GB, Sundararajan, N, and Saratchandran, P (2009). Online sequential fuzzy extreme learning machine for function approximation and classification problems. IEEE Transactions on Systems, Man, and Cybernetics Part B (Cybernetics). 39, 1067-1072. https://doi.org/10.1109/TSMCB.2008.2010506
    Pubmed CrossRef
  31. Ontiveros-Robles, E, Castillo, O, and Melin, P (2020). An approach for non-singleton generalized type-2 fuzzy classifiers. Journal of Intelligent & Fuzzy Systems. 35, 7203-7215. https://doi.org/10.3233/JIFS-200639
    CrossRef
  32. Ontiveros, E, Melin, P, and Castillo, O (2020). Designing hybrid classifiers based on general type-2 fuzzy logic and support vector machines. Soft Computing. 24, 18009-18019. https://doi.org/10.1007/s00500-020-05052-x
    CrossRef
  33. Rubio, E, Castillo, O, Valdez, F, Melin, P, Gonzalez, CI, and Martinez, G (2017). An extension of the fuzzy possibilistic clustering algorithm using type-2 fuzzy logic techniques. Advances in Fuzzy Systems. 2017. article no 7094046. https://doi.org/10.1155/2017/7094046
    CrossRef
  34. Huang, GB, Chen, L, and Siew, CK (2006). Universal approximation using incremental constructive feedforward networks with random hidden nodes. IEEE Transactions on Neural Networks. 17, 879-892.
    Pubmed CrossRef
  35. Huang, GB, and Chen, L (2007). Convex incremental extreme learning machine. Neurocomputing. 70, 3056-3062. https://doi.org/10.1016/j.neucom.2007.02.009
    CrossRef
  36. Gepperth, A, and Hammer, B . Incremental learning algorithms and applications., Proceedings of European Symposium on Artificial Neural Networks (ESANN), 2016, Bruges, Belgium.
  37. Bermejo, P, de La Ossa, L, Gamez, JA, and Puerta, JM (2011). Enhancing incremental feature subset selection in high-dimensional databases by adding a backward step. Computer and Information Sciences II. London, UK: Springer, pp. 93-97 https://doi.org/10.1007/978-1-4471-2155-8_11
    CrossRef
  38. Perkins, S, Lacker, K, and Theiler, J (2003). Grafting: fast, incremental feature selection by gradient descent in function space. Journal of Machine Learning Research. 3, 1333-1356.
  39. Bermejo, P, Gamez, J, and Puerta, J . On incremental wrapper-based attribute selection: experimental analysis of the relevance criteria., Proceedings of the 12th International Conference Information Processing and Management of Uncertainty for Knowledge-Based Systems (IPMU), 2008, Torremolinos, Malaga, Spain, pp.638-645.
  40. Ruiz, R, Riquelme, JC, and Aguilar-Ruiz, JS (2006). Incremental wrapper-based gene selection from microarray data for cancer classification. Pattern Recognition. 39, 2383-2392. https://doi.org/10.1016/j.patcog.2005.11.001
    CrossRef
  41. d’Alche-Buc, F, and Ralaivola, L (2002). Incremental learning algorithms for classification and regression: Local strategies. AIP Conference Proceedings. 627, 320-329. https://doi.org/10.1063/1.1503700
    CrossRef
  42. Feng, G, Huang, GB, Lin, Q, and Gay, R (2009). Error minimized extreme learning machine with growth of hidden nodes and incremental learning. IEEE Transactions on Neural Networks. 20, 1352-1357. https://doi.org/10.1109/TNN.2009.2024147
    Pubmed CrossRef
  43. Yang, Y, Wang, Y, and Yuan, X (2012). Bidirectional extreme learning machine for regression problem and its learning effectiveness. IEEE Transactions on Neural Networks and Learning Systems. 23, 1498-1505. https://doi.org/10.1109/TNNLS.2012.2202289
    Pubmed CrossRef
  44. Baluja, S (1994). Population-based incremental learning: a method for integrating genetic search based function optimization and competitive learning. Pittsburgh, PA, USA: Carnegie Mellon University Technical Report No. CMU-CS-94-163
  45. Wang, A, An, N, Chen, G, Yang, J, Li, L, and Alterovitz, G . Incremental wrapper based gene selection with Markov blanket., Proceedings of 2014 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), 2014, Belfast, UK, Array, pp.74-79. https://doi.org/10.1109/BIBM.2014.6999251
    CrossRef
  46. Kale, A, and Sonavane, S (2019). F-WSS++: incremental wrapper subset selection algorithm for fuzzy extreme learning machine. International Journal of Machine Learning and Cybernetics. 10, 1821-1832. https://doi.org/10.1007/s13042-018-0859-9
    CrossRef
  47. Wang, S, Ouyang, L, Yuan, Y, Ni, X, Han, X, and Wang, FY (2019). Blockchain-enabled smart contracts: architecture, applications, and future trends. IEEE Transactions on Systems, Man, and Cybernetics: Systems. 49, 2266-2277. https://doi.org/10.1109/TSMC.2019.2895123
    CrossRef
  48. UC Irvine Machine Learning Repository. [online] Available: http://archive.ics.uci.edu/
  49. Sun, Y, and Li, J . Iterative relief for feature weighting., Proceedings of the 23rd International Conference on Machine Learning, 2006, Pittsburgh, PA, USA, Array, pp.913-920. https://doi.org/10.1145/1143844.1143959
    Pubmed CrossRef
  50. Hall, MA, and Holmes, G (2003). Benchmarking attribute selection techniques for discrete class data mining. IEEE Transactions on Knowledge and Data Engineering. 15, 1437-1447. https://doi.org/10.1109/TKDE.2003.1245283
    CrossRef
  51. Urbanowicz, RJ, Meeker, M, La Cava, W, Olson, RS, and Moore, JH (2018). Relief-based feature selection: introduction and review. Journal of Biomedical Informatics. 85, 189-203. https://doi.org/10.1016/j.jbi.2018.07.014
    Pubmed KoreaMed CrossRef
  52. Nahato, KB, Nehemiah, KH, and Kannan, A (2016). Hybrid approach using fuzzy sets and extreme learning machine for classifying clinical datasets. Informatics in Medicine Unlocked. 2, 1-11. https://doi.org/10.1016/j.imu.2016.01.001
    CrossRef
  53. Mahdiyah, U, Irawan, MI, and Imah, EM (2015). Integrating data selection and extreme learning machine for imbalanced data. Procedia Computer Science. 59, 221-229. https://doi.org/10.1016/j.procs.2015.07.561
    CrossRef
  54. Subbulakshmi, C, and Deepa, S (2015). Medical dataset classification: a machine learning paradigm integrating particle swarm optimization with extreme learning machine classifier. The Scientific World Journal. 2015. article no 418060. https://doi.org/10.1155/2015/418060
    Pubmed KoreaMed CrossRef
  55. Novakovic, J, Strbac, P, and Bulatovic, D (2011). Toward optimal feature selection using ranking methods and classification algorithms. Yugoslav Journal of Operations Research. 21, 119-135. https://doi.org/10.2298/YJOR1101119N
    CrossRef
  56. Wuensch, KL (1996). Straightforward statistics for the behavioral sciences. Journal of the American Statistical Association. 91, 1750-1751. https://doi.org/10.2307/2291607
    CrossRef
  57. Devaraj, S, and Paulraj, S (2015). An efficient feature subset selection algorithm for classification of multidimensional dataset. The Scientific World Journal. 2015. article no 821798. https://doi.org/10.1155/2015/821798
    Pubmed KoreaMed CrossRef
  58. Kothari, CR (2004). Research Methodology: Methods and Techniques. New Delhi, India: New Age International
  59. Wong, SY, Yap, KS, Yap, HJ, Tan, SC, and Chang, SW (2015). On equivalence of FIS and ELM for interpretable rule-based knowledge representation. IEEE Transactions on Neural Networks and Learning Systems. 26, 1417-1430. https://doi.org/10.1109/TNNLS.2014.2341655
    Pubmed CrossRef
  60. Quteishat, A, and Lim, CP (2008). A modified fuzzy min-max neural network with rule extraction and its application to fault detection and classification. Applied Soft Computing. 8, 985-995. https://doi.org/10.1016/j.asoc.2007.07.013
    CrossRef
  61. Yap, KS, Lim, CP, and Mohamad-Saleh, J (2010). An enhanced generalized adaptive resonance theory neural network and its application to medical pattern classification. Journal of Intelligent & Fuzzy Systems. 21, 65-78. https://doi.org/10.3233/IFS-2010-0436
    CrossRef
  62. Calisir, D, and Dogantekin, E (2011). An automatic diabetes diagnosis system based on lda-wavelet support vector machine classifier. Expert Systems with Applications. 38, 8311-8315. https://doi.org/10.1016/j.eswa.2011.01.017
    CrossRef
  63. Jaganathan, P, and Kuppuchamy, R (2013). A threshold fuzzy entropy based feature selection for medical database classification. Computers in Biology and Medicine. 43, 2222-2229. https://doi.org/10.1016/j.compbiomed.2013.10.016
    Pubmed CrossRef
  64. Temurtas, H, Yumusak, N, and Temurtas, F (2009). A comparative study on diabetes disease diagnosis using neural networks. Expert Systems with Applications. 36, 8610-8615. https://doi.org/10.1016/j.eswa.2008.10.032
    CrossRef
  65. Polat, K, Gunes, S, and Arslan, A (2008). A cascade learning system for classification of diabetes disease: generalized discriminant analysis and least square support vector machine. Expert Systems with Applications. 34, 482-487. https://doi.org/10.1016/j.eswa.2006.09.012
    CrossRef
  66. Aslam, MW, Zhu, Z, and Nandi, AK (2013). Feature generation using genetic programming with comparative partner selection for diabetes classification. Expert Systems with Applications. 40, 5402-5412. https://doi.org/10.1016/j.eswa.2013.04.003
    CrossRef
  67. Varma, KV, Rao, AA, Lakshmi, TSM, and Rao, PN (2014). A computational intelligence approach for a better diagnosis of diabetic patients. Computers & Electrical Engineering. 40, 1758-1765. https://doi.org/10.1016/j.compeleceng.2013.07.003
    CrossRef
  68. Mattila, J, Koikkalainen, J, Virkki, A, Van Gils, M, and Lotjonen, J (2012). Design and application of a generic clinical decision support system for multiscale data. IEEE Transactions on Biomedical Engineering. 59, 234-240. https://doi.org/10.1109/TBME.2011.2170986
    Pubmed KoreaMed CrossRef
  69. Martinez-Estudillo, FJ, Hervas-Martinez, C, Gutierrez, PA, and Martinez-Estudillo, AC (2008). Evolutionary product-unit neural networks classifiers. Neurocomputing. 72, 548-561. https://doi.org/10.1016/j.neucom.2007.11.019
    CrossRef
  70. Hervas-Martinez, C, Martinez-Estudillo, FJ, and Carbonero-Ruz, M (2008). Multilogistic regression by means of evolutionary product-unit neural networks. Neural Networks. 21, 951-961. https://doi.org/10.1016/j.neunet.2007.12.052
    Pubmed CrossRef
  71. Zhang, X, and Zhang, F . Images features extraction of tobacco leaves., Proceedings of 2008 Congress on Image and Signal Processing, 2008, Sanya, China, Array, pp.773-776. https://doi.org/10.1109/CISP.2008.88
    CrossRef
  72. Ismail, N, Othman, ZA, and Samsudin, NA (2019). Regularization activation function for extreme learning machine. International Journal of Advanced Computer Science and Applications. 10, 241-248. https://doi.org/10.14569/ijacsa.2019.0100331
    CrossRef
  73. Mi, Y, Yuan, S, Li, X, and Zhou, J (2021). Dense residual generative adversarial network for rapid rain removal. IEEE Access. 9, 24848-24858. https://doi.org/10.1109/ACCESS.2021.3055527
    CrossRef
  74. Garcia-Gasulla, D, Pares, F, Vilalta, A, Moreno, J, Ayguadé, E, Labarta, J, Cortes, U, and Suzumura, T (2018). On the behavior of convolutional nets for feature extraction. Journal of Artificial Intelligence Research. 61, 563-592. https://doi.org/10.1613/jair.5756
    CrossRef

Archana P. Kale is working as an associate professor, Modern Education Society’s College of Engineering, Savitribai Phule Pune University, with the Department of Computer Engineering, Pune, Maharashtra, India. She completed the Ph.D. degree in Computer Science and Engineering from Walchand College of Engineering, Shivaji University, Kolhapur, Maharashtra, India. She has published research paper in Springer journal and conference including IEEE and it’s also available online. Her main research work focuses on artificial intelligence, machine learning, deep learning, data analytics, Internet of Things, fuzzy logic and computational intelligence. She has 21 years of teaching experience and 3 years of research experience.

Article

Original Article

International Journal of Fuzzy Logic and Intelligent Systems 2023; 23(4): 465-481

Published online December 25, 2023 https://doi.org/10.5391/IJFIS.2023.23.4.465

Copyright © The Korean Institute of Intelligent Systems.

Data Analysis for Fuzzy Extreme Learning Machine

Archana P. Kale

Modern Education Society’s College of Engineering, Savitribai Phule Pune University, Maharashtra, India

Correspondence to:Archana P. Kale (archana.mahantakale@gmail.com)

Received: August 17, 2022; Accepted: December 22, 2023

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/3.0/) which permits unrestricted noncommercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract

Fuzzy extreme learning machine (F-ELM) one of the learning algorithm which is specifically used for uncertainty classification. Uncertainty classification is critical problem in the area of machine learning. In various real-time applications, the ambiguity is present in the input dataset itself which affects the systems generalization performance. Data (feature) analysis plays a major role in such types of problems. To solve the said problem in this paper, distance-based Relief algorithm and fuzzy extreme learning algorithm are used for data analysis and classification, respectively which contributes to Relief-based data analysis for F-ELM (RFELM++) and Relief-based data analysis for online sequential ELM (RFOSELM++) for batch mode and sequential input, respectively. Experimental results are calculated by using clinical dataset. Through the results, it is observed that RFELM++ produces increased accuracy in comparison with RELM++ for clinical dataset. The RFOSELM++ maintain accuracy by using 41.5% features as differentiate to OS-ELM for the UCI Repository dataset. The proposed RFOSELM++ algorithm is compared with similar already available sequential based algorithms. As a case study, a novel application of classification of nutrient deficiency and plant disease in which ambiguity presents is considered. Both proposed algorithms are exploited in this expert system which helps the remote farmer with expert advice. The proposed RFELM++ algorithm is tested and validated by using statistical methods.

Keywords: Uncertainty classification problem, Sequential problem, Internet of Things, Feature subset selection problem, Pattern classification

1. Introduction

Classification, feature subset selection (FSS), and sequential problems are some of the critical problems in machine learning which has drawn substantial aspect from researchers in recent years. Due to the technological advances, data gets generated at an ever increasing space and the dimensionality or size of the datasets are increased by the day. Therefore, it is mandatory to design and develop the potent and adequate machine learning methods, that can be used to select, analyse and extract only the beneficial features. The existence of the dispensable and immaterial features in input may decreases the performance of systems. FSS is critical important pre-processing technique which selects only prominent (non-redundant and relevant) features. Pattern classification is a critical process which is utilize to classify the input dataset into one of the output classes.

As database and computer technologies advance speedily, various advancing challenges appear with respect to FSS and classification problems as follows:

  • • Improvement in accuracy by using the same number of attributes;

  • • Maintain accuracy by downsizing the attributes;

  • • Improvement in accuracy by reducing the number of attributes.

FSS problem has drawn considerable attention from researchers in current years due to the existence of a huge attributes. Many real-time applications contain high-dimensional dataset. These thousands of features are usually provided to the learning algorithms for the classification task.

In many real-time applications, the attributes present in the dataset are ambiguous. Therefore, it becomes very challenging to classify the instance of input. Such kind of the problems in which the ambiguity presents in the attributes of the input dataset [1] are come under the uncertainty classification problem (UCP). Uncertainty classification is used for one class classification [2], binary classification problems [3,4] and imbalance classification problem via support vector machine (SVM) [5]. A gradient approach is developed in Naive Bayes (NB) for value uncertainty classification learning by Lee [6]. Zhang and Ji [1] developed fuzzy extreme learning machine (F-ELM) for uncertainty classification problem.

F-ELM is the latest classifier which associates the working of fuzzy inference system (FIS) and ELM. For uncertainty classification problem F-ELM algorithm is mostly used [1, 7]. F-ELM uses a membership function for required input which is represented by linguistic variables.

Many research papers show the different FSS techniques for various classification problems. However, FSS with incremental learning for uncertainty classification problem is absent in the vast literature survey.

The original F-ELM and ELM are the batch learning mode which considers that the data is available initially. So, F-ELM and ELM classifiers are unable to handle sequent or serial input. So that, online sequential extreme learning machine (OS-ELM) algorithm is presented by Nan-Ying et. al. for sequent input. Three layers are available for ELM and OS-ELM: input, hidden, and output. Randomly initialization of weights is carried out from the first (input) to second (hidden) layer and logically modulated the weights from the second (hidden) to third (output) layer. On account of random initialization, generalized performance of system may degrade. So that it is necessary to select the prominent features.

To distinguish said problem, in this paper incremental F-ELM (RFELM++) and incremental fuzzy online sequential ELM (RFOSELM++) for the UCP are proposed for batch mode and sequential input respectively. Thus, the key intent of this paper is to design RFELM++ and RFOSELM++ algorithm by using the activation function like radial basis function (RBF) and additive function which is missing in the vast literature survey. Another aspect of the paper is to design an IoT based expert system - Precision Agriculture by exploiting the proposed algorithms.

The overall structure of the paper is the following: in Section 2, the literature survey is detailed in the related work. The proposed incremental F-ELM and RFOSELM++ algorithms are depicted in Section 3. The experimental observations and results of clinical datasets are discussed and compared in Section 4. The validation and testing part of the proposed algorithm are discussed in detail in Section 5 as the results and discussions. Section 6 outlined the conclusion with further research direction.

2. Related Work

2.1 Feature Subset Selection

In various growing areas FSS plays a very critical task. The key role of FSS is to choose prominent/optimal (relevant and non-redundant) feature subset which produces similar or improved (in some of the cases) results as compared to the prime or base data set in which all features are present. It is very hard to decide the significance of attributes without preparatory information [8]. Therefore, FSS is an prime step to select prominent features for efficient machine learning applications [9, 10]. The primary intention of using any FSS techniques is for [11]:

  • • Improvement in classification accuracy,

  • • Improvement in prediction performance,

  • • Improvement in model interpretability,

  • • Dimensionality reduction,

  • • Storage management.

2.1.1 FSS techniques

The few of the FSS techniques which are mostly used are: filter, wrapper, and hybrid.

Filter techniques rank the features or select feature subset independently of the learning algorithm. Filter techniques are easy to interpret and fast. The characteristics of filter techniques are considered independently, totally independent of which learning algorithm is used etc. For class separability, the evaluation criteria is required which is a classical criterion of the filtering present in the literature. The criteria like distance, information and reliant are mostly utilize to classify the features.

Wrapper techniques use a learning algorithm for FSS, training is required for every feature subset. So, wrapper techniques are computationally heavy and dependent on the type of classifier used. The characteristics of wrapper techniques are computationally expensive, use only heuristic search and dependent on the learning algorithm.

Hybrid techniques use both filter and wrapper techniques. Ranking-wrapper techniques [12], Relief-F with sequential backward search [13] and boosting-based hybrid for feature selection (BBHFS) [14] are few of hybrid techniques.

Tahir and Loo [15] used Relief-F method for feature selection in continual learning framework. Various optimization algorithms like particle swarm intelligence [16] is used to solve feature selection problem and multi-objective particle swarm optimization algorithm [17] is used to solve bioprocess application problems and tumor treatment problems.

2.2 Pattern Classification

Pattern classification is specially used for decision making task. It classifies input entity to one of the predetermined target outputs. A supervised classification is specially used to categorised the given input data into one of the present target classes [18]. Having the data D = xi, yi, where i = 1, …, K and xiRn is a single pattern or instance with n features, yi ∈ {1, 2, …, m}. yi is the total number of target class which is represented by m and K represents total number of patterns (D); a problem of pattern classification is the determination of a mapping model M(.), where M(xi) = yi [19]. Neural network (NN) is a robust computational tool which is specifically used for classification. Various NN based classifiers–backpropagation NN (BPNN), multilayer perceptron (MLP), radial basis function (RBF), decision tree, support vector machine (SVM), C4.5, fuzzy NN (FNN), convolutional neural network (CNN), ELM, etc.–are required for classification. Tao et al. [20] designed a deep CNN for detecting the defects in insulators. Wen et al. [21] developed a deep transfer learning (DTL) method for fault diagnosis. ELM is used in breast cancer detection with the help of multilayer fuzzy expert system which was developed by Mojrian et al. [22]. Basically, the classifier plays a very crucial act in system overall performance. The CNN classifier faces various problems like over fitting problem, local maxima and minima problem, etc. Therefore, classifier selection is a challenging task.

2.2.1 ELM

The original ELM classifier is basically designed for a batch mode. ELM is acknowledged that the complete input data is available previously. However, many real applications require to handle the data sequentially. Therefore, Nan-Ying Liang et. al. have designed OS-ELM algorithm for handling the sequent or serial input. OS-ELM randomly assigns all the weight values in-between first (input) and second (hidden) layer and the weight values in between second (hidden) to third (output layer) are analytically tuned [2325] as demonstrated in Figure 1 [26]. The global generalization performance may degrade due to the existence of non-prominent attributes and random initialization. Therefore, it is necessary to choose a prominent feature subset for sequential input. However, ELM is insufficient to provide a solution for the sequential and UCP.

2.2.2 OS-ELM

OS-ELM is exploited in various applications like estimation of hematocrit [27], DWT domain watermarking [28], traffic profiling [29], etc. However, the literature survey lacks the exploitation of incremental learning ability by using prominent or optimal feature subset selection for sequent or serial input.

2.2.3 F-ELM

Traditional ELM is unable to solve some real-time time problems like the UCP and imbalance problem [1]. So, F-ELM is designed by Zhang and Ji [1] to solve specially UCP. However, in order to solve the UCP for sequential input OS-ELM is designed [30].

Ontiveros-Robles et al. [31] used type II fuzzy classification for non-singleton and also designed hybrid classification by using SVM [32]. Rubio et al. [33] used fuzzy clustering for unsuprvised method.

2.3 Incremental Learning

Incremental learning is used in various ways like incrementally add hidden nodes [34, 35], incrementally add streaming data [36], incremental FSS [3740], etc. The exploitation of incremental learning in terms of incremental FSS for sequential input is also missing in the literature survey.

Incremental learning comes in various forms and definitions as

  • • Incrementally add streaming data: Incremental learning is a learning for the data which appear sequentially or over time [36, 41]. Alexander Gepperth describes online learning, incremental learning and concept drift for the supervised learning paradigm in detail.

  • • Incrementally add hidden nodes: Incremental learning defines as an incrementally insertion of the new neurons to the hidden layer. Error-minimized ELM (EM-ELM) [42], Convex incremental ELM (CI-ELM) [35] and uncremental extreme learning machine (RELM++) [34] are some of the examples of incremental learning. Yang et al. [43] designed a bidirectional extreme learning machines (B-ELM) algorithm for regression.

  • • Incremental feature subset selection: Incremental learning in the form of incremental feature subset selection is used in many papers like [3740]. Baluja [44] have developed a population-based incremental learning method which integrated learning and the random search based function optimization [44]. Tahir and Loo [15] used an incremental ELM is used for food recognition. Wang et al. [45] have designed incremental wrapper based gene selection with Markov blanket. Incremental wrapper subset selection (IWSS) uses a filter measure for feature ranking by using a sequential search [46]. Then, a SFS is used for incrementally insertion of the features.

Wang et al. [47] designed a framework for smart contracts based on a novel six-layer architecture.

3. Proposed Methodology

This section is categorized into two subsections, RFELM++ and RFOSELM++ algorithms.

3.1 RFELM++ Algorithm

In this section, incremental learning is used to resolve the uncertainty classification problem. As already discussed in Section 2, F-ELM specifically uses for the uncertainty classification problem. Hence, RFELM++ algorithm is initiated. The paradigm of RFELM++ is broadly categorised into four subsystems: input, incremental learning for FSS, fuzzification, and classification.

Input subsystem

In this section, the detailed description of the input is described in detail. In order to resolve the UCP, it is necessary to use the datasets in which uncertainty is present. Therefore, Statlog Heart Disease (SHD) and Pima Indian Diabetes (PID) datasets are used [48]. The PID data set contains total 768 instances and 8 attributes. The datasets consists of several medical predictor variables and one target variable, outcome. Predictor variables includes the number of pregnancies the patient has had, their body mass index (BMI), insulin level, age, and so on. The SHD data set contains 270 instances and 13 attributes. Additionaly, the Cleveland Heart Disease (CHD) dataset is used for multiclass classification.

Incremental learning for FSS

The generalization performance of any system is depend on the input. If the input to the system is right the performance of system gets increased. Therefore, it is very important task to provide a right or correct input features to the system.

Prominent feature analysis

In data preprocessing FSS algorithm is used. To select only prominent features relief algorithm is used. Relief algorithm is one of the FSS algorithms based on distance measure. The measure of distance (D) is basically used to search the features which can differentiated between two target classes. If the D is equal to 0, then the two attributes are identical. The Relief algorithm is used the joint relationship to compute weights with the output [49, 50] which is famous due to it’s simplicity and effectiveness. In Relief (feature weight-based algorithm), a statistical method is used to choose the relevant attributes [51].

After applying relief algorithm on PID and SHD datasets, we get feature order. For example, PID dataset has 8 attributes like age, DPF, BMI, insulin, skin, BP, Glu, and pregnancy. After Relief algorithm we get ⟨2, 6, 4, 8, 1, 7, 5, 3⟩ feature order. With this order it is understood that 2 is the most required or more relevant feature whereas 3 is the least required feature.

As per already described in the previous section, prominent feature order is calculated by considering distance measures i.e., Relief algorithm. Suppose the attribute order for SHD and PID data sets is ⟨13, 12, 3, 9, 10, 11, 2, 1, 7, 4, 5, 6, 8⟩ and ⟨2, 6, 8, 1, 7, 5, 4, 3⟩ respectively. The organised attributes are incrementally inserted one by one by using SFS for sequential space. Experimental observation shows that from all the subsets {1, 2, 3, 4, 5, 7, 9, 10, 11, 12, 13} and {1, 2, 4, 5, 6, 7, 8} are the prominent feature subset for SHD and PID datasets, respectively, as it provides improved classification accuracy.

Fuzzification and classification

The attributes of PID and SHD datasets are translated into linguistic attributes (fuzzified features) with the help of trapezoidal membership function [52]. The PID (8 attributes) and SHD (13 attributes) datasets are translated into 25 and 39 linguistic attributes, respectively. Input dataset is distributed into two datasets like training with 70% and testing with 30 %. Hence for PID ans SHD datasets 577 and 177 instances are used for traing and 577 and 177 instances are used for testing, respectively. By using trapazoidal membership function the features are converted into fuzzified features. For classification ELM is used.

3.2 RFOSELM++ Algorithm

ELM is a batch learning mode classifier which is not suitable for Sequential input. In this section, incremental learning for FSS is used to solve the UCP as well as sequential problem. Hence, incremental FSS techniques for the UCP is proposed for sequential input which contributes to RFOSELM++ algorithm.

In RFOSELM++, the ranked features are inserted sequentially i.e., one by one. The accuracy is calculated for all feature subsets which are equal to the number of attributes. The proposed RFOSELM++ learning algorithm is applied for every subset and predicted accuracy is calculated. The paradigm of RFOSELM++ is mostly same as RFELM++. Only the classification subsystem is different one. Through observations, the prominent feature subset is one which provides maximum accuracy.

4. Experimental Results

For experimental observation, Waikato Environment for Knowledge Analysis (WEKA) and MATLAB R2014a are used. Precision, Recall and classification accuracy are used as an evaluation measure. The calculations of evaluation measures are depend upon the contingency table or confusion matrix of classifier. Confusion matrix contains both true and false positives (TP/FP) also true and false negatives (TN/FN) [52]. FP correspond to negative samples falsely identified as positive. FN correspond to positive samples falsely resembles as negative. TP are the samples perfectly identified as positive. TN refer to negative samples perfectly identified as negative. The perfectness of identified examples are measured by using precision. The measure of the all positive identified examples is called recall [53]. Accuracy, precision and recall (sensitivity) are measured by evaluating the respective values in Eqs. (1) to (3), respectively

Acc=TP+TNTP+FP+TN+FN,Pre=TPTP+FP,Rec=TPTP+FN.

In this section, the incremental learning capability for batch and sequential mode is exploited i.e., incremental learning for F-ELM (RFELM++) algorithm and incremental learning for fuzzy online sequential extreme learning machine, respectively.

4.1 RFELM++ (Prominent Feature Subset)

The experimental results of the proposed RFELM++ algorithm are enumerated. In order to prove the effectiveness of an initiated algorithm the same steps are implemented for ELM classifier (RELM++). The results of both RELM++ and RFELM++ are compared with sigmoidal activity function by considering PID and SHD datasets. The overall comparison of RELM++ and RFELM++ for PID dataset is as shown in Figure 2. The same results are evaluated for SHD dataset also. Through the results, it is observed that RFELM++ produces 6.266% and 0.855% increased accuracy in comparison of RELM++ for PID and SHD datasets consequently.

The overall performance of the FELM classifier is analysed by considering the precision, accuracy and recall. Figure 3 shows an overall comparison of performance for ELM and F-ELM with all features. The results are calculated by using all activity functions like sigmoid function (sig), radial basis function (rbf), sine function (sin), hardlim function (hardlim), triangular basis function (tribas) for SHD and PID datasets. By observations it is apprise that, F-ELM issues an improved performance as compared to ELM classifier.

Various FSS methods like half selection (HS), mean selection (MS), particle swarm optimization (PSO) based FS, neural network for threshold, self-regulated learning (SR-PSO) [54] are used to select prominent features. Table 1 indicates the output of all given FSS methods with ELM and FELM classifiers. For evaluation, the sigmoidal activation function is used for PID dataset. From experimented results, it is noticed that the F-ELM achieves 4.65% improved classification accuracy in comparison with ELM for the same number of feature. The result for RFELM++ algorithm for multiclass classification problem for CHD dataset is as shown in Figure 5.

In order to prove the effectiveness of the proposed RFELM++ algorithm, experiment results are derived and analysed by using the traditional classifiers–MLP, random forest (RF), Bayes net (BN), SVM, NB, RBF, and J4.8–by using prominent features. The prominent features are selected by using Relief algorithm. Five-fold cross-validation method is used for evaluation. Figure 4 shows the variation of the proposed RFELM++ algorithm with existing classifiers by using relief FSS algorithm for PID and SHD data set. It is evidently prove that the proposed RFELM++ algorithm enables maximum predictive accuracy with minimum number of features.

4.2 RFOSELM++ (Prominent Feature Subset)

PID dataset is used as an input dataset. Normalization is used in pre-process. The algorithm uses relief algorithm (distance measure) for ranking the features. IWSS [39, 40] is used which acquires a ranking of the features over to the target class. IWSS is used SFS search over the ranked attributes by incrementally inserting those features one by one. The wrapper way is used to measure the relevance of the new attribute. The classification accuracy is calculated by using OS-ELM classifier.

Experimental results of incremental attribute learning for sequential input is calculated for both ELM and F-ELM classifier. Table 2 shows the parameters required for the evaluation. Figures 6 and 7 show the comparison of RELM++ and RFELM++ for training and testing accuracy, respectively. Through observation, it is noticed that the proposed RFELM++ issues an improved as well as stable (constant) results as compared to RELM++.

5. Result and Discussions

In order to prove the validation of the proposed RFELM++ algorithm, experimental observations need to be validated and hypothetically tested [55]. F-ELM classifier is used after the feature selection by using statistical methods. The PID and SHD datasets are categorized into four samples like S4 (SHD ELM), S3 (PID ELM), S2 (SHD F-ELM) and S1 (PID F-ELM). The comparative analysis of before and after relief feature selection algorithm is as shown in Figure 8. Through the results it is noticed that, RFELM++ algorithm with feature subset produces improved performance as compared to with all features.

The correlation analysis technique is used to compute the robustness of the relationship between 2 or more variables. All the time the values of correlation are in between −1 to +1. The abbreviation of +1 is that the two variables are strongly related with positive linear, −1 is that two variables are resolutely related with negative linear and 0 is that no relationship in between the two variables. According to Evans classification, the correlation coefficient are broadly categorized into very strong, strong, moderate, weak, very weak [56].

The c (Pearson’s correlation coefficient) is given by

cxy=pq[n.p2-(p)2].[n.q2-(q)2],

where ‘p’ is metric before feature selection and ‘q’ is metric after feature selection.

The coefficient (c = 0.6177) is calculated by using all features and Relief FS. Figure 8 illustrates the computation of evaluation metrics accuracy for ELM and F-ELM learning algorithms. Through this computation, it proves that the relationship in between the F-ELM (all features) and RFELM++ (feature subset) is strong as per Evans classification [56].

When the relationship (c) is strong, then the Student’s paired t-test is utilised for comparison in-between two discrete techniques. Paired t-test (pt-value) is calculated by using Eq. (5). To check the competence of the initiated RFELM++ algorithm, Student’s paired t-test is used. Based on the value of r = 0.6177, the t-value is calculated, i.e., −1.097.

pt=qq2-(q)2/n-1,

where ‘q’ is the difference between ‘a’ (before applying FSS algorithm) and ‘b’ (after applying FSS algorithm) (qi = aibi).

Hypothesis of testing [57] RFELM++:

  • H0: No variance between the generalization performance of F-ELM (with all features or before feature selection) and RFELM++ (after feature selection)

  • H1: Variance between the generalization performance of F-ELM (with all features or before feature selection) and RFELM++ (after feature selection)

The pt value i.e., Student’s t-distribution for the degree of freedom 4 is (2.776, ∝= 0.05) and (4.604, ∝= 0.01) [58]. By observation, it is notified that there is no variance between the generalization performance of F-ELM (with all features or before feature selection) and RFELM++ (after feature selection) for four samples, where ∝ is the level of significance. The main reasons behind the improved performances of the proposed algorithm are the removal of irrelevant and redundant features from the input dataset before weighted classification.

5.1 Benchmark Comparison

For performance comparison, accuracy of RFELM++ by using prominent feature subset is compared with F-ELM [59], modified fuzzy min-max neural network-FIS (MFMM-FIS) [60] and enhanced generalized adaptive resonance theory FIS (EGART-FIS) [61] for the PID dataset. Table 3 describes the comparative classification accuracy for the same. With the comparative analysis, it is observed that the testing and training accuracy is 83.47% to 86.80%, respectively by using only 62.5% features.

5.1.1 RFELM++ and the existing classifiers with different FSS methods

With the intention to evince the effectiveness of the proposed RFELM++ algorithm, it is mandatory to have comparison the result of RFELM++ with the similar approach like NN for threshold selection (NN) [62,63], multilayer NN (MLNN) by using Levenberg–Marquardt (LM) [64], generalized discriminant analysis least square (GDA-LS) [65], HS method, least square-ELM (LS-ELM) [65], linear discriminant analysis and Morlet wavelet (LDA-MW), probabilistic neural network (PNN) [64], MS methods. Table 4 shows RFELM++ and other classifiers comparison for PID input dataset. Table 5 illustrates the same comparison for SHD dataset with evolutionary product-unit NN (EPUNN), MS, evolutionary sigmoidal unit NN (ESUNN), HS, multilogistic regression by means of evolutionary product-unit NN (MR+EPUNN) for SHD dataset. Through comparative average analysis, it is noticed that the introduced RFELM++ issues 8.82% and 6.63% increased accuracy by reducing 6.41% and 10.71% for SHD adn PID datasets, respectively.

The comparative analysis in between F-ELM and ELM by using different FSS algorithms is as shown in Figure 9 for PID dataset. HS, PSO, MS, NN, T-test and self-regulated learning (SR-PSO) [54] are used as FSS algorithms. For PID dataset, the comparison of ELM and F-ELM classifier in terms of accuracy measure by using the feature subset as shown in Figure 9 in which F-ELM produces increased performance for the same number of features.

The introduced RFELM++ algorithm is exploited in the real time application which is represented as a case study in the following section.

6. IoT-based Precision Agriculture Expert System

Agriculture is the backbone of Indian economy where approximate 60% of people are dependent directly or indirectly on agriculture. The expert advice is required for distinguishing the plant disease damage and nutrient imbalance. It is observed that, the conventional judgmental analysis is not enough while deciding the quantity of chemical or fertilizer to be used. The mis-proportional dose harms directly the health of the crop and hence the living beings. To overcome the said problem, this section is developed an IoT-based precision agriculture expert system by inferring proposed algorithms with improving the accuracy by reducing the features or by selecting the prominent features. Grape leaf diseases like downy mildew and nutrient deficiency which contains ambiguity is considered as a case study. Both are spread with yellow spot on leaf. The conceptual framework for the proposed IoT-based precision agriculture expert system is as shown in Figure 10.

In IoT-based precision agriculture expert system, the system is partitioned into three rough phases namely image processing, FSS by using incremental learning and classification. In image processing, the acquisition step is used to input the image. The image is acquired by using cell phone or any digital camera. The resolution of the image is 96 dpi (dots per inch). For interfacing, the ARDUINO hardware is used with various sensors like temperature, humidity etc. The leaf image is used to extract the feature points and are stored in the database.

Data normalization is used as the preprocessing method. In image analysis, image segmentation and feature extraction methods are considered. For image segmentation, region-based segmentation is used to separate the healthy and diseased region by using leaf color. The features extracted from an images are color, shape and texture [71]. Once features are extracted, the database is generated with integrated information from temperature and humidity sensors. On that created dataset the proposed algorithms are applied for feature selection and classification.

In order to increase the generalization performance of the system, the proposed RELM++ and RFOSELM++ algorithms are exploited. The performance of precision agriculture is evaluated and tested by using RELM++ and RFOSELM++ algorithms. The comparison in-between RELM++ and ELM for proposed system (plant disease and nutrient deficiency) is as shown in Figure 11 in terms of testing accuracy. Through this experimentation, it is intended that the RELM++ produces 5.71% increased accuracy and RFOSELM++ maintains (same accuracy) the accuracy with only 27.27% features. However, for Tribas activation function the classical method provides good results as compared to the proposed one. Tribas provided an optimal result only for certain number of hidden neurons for ELM [72].

7. Conclusion

The prime intention of this paper is to exploit the two proposed algorithms like RFELM++ and RFOSELM++ into the new developed IoT-based precision agriculture Expert system. The expert system helps farmers in decision making. The experimental results for ELM, OS-ELM, RELM++, RFOSELM++, RFELM++ and RFOSELM++ are calculated. RFELM++ produces 6.266% and 0.855% increased accuracy in comparison of RELM++ for PID and SHD datasets consequently. Additionally, various FSS methods like like HS, MS, PSO based FS, neural network for threshold, self-regulated learning (SR-PSO) [54] are used to select prominent features. Through the experimented results, it is noticed that the F-ELM achieves 4.65% improved classification accuracy in comparison with ELM for the same number of feature. The proposed RFOSELM++ maintain the accuracy by using 41.5% features as differentiate to OS-ELM for UCI Repository dataset. For benchmark problem, the results of initiated RFELM++ algorithm is compared with the similar existing approaches. From the comparative average analysis, it is noticed that the introduced RFELM++ issues 8.82% and 6.63% increased accuracy by reducing 6.41% and 10.71% for SHD and PID datasets, respectively. Table 3 describes the comparative classification accuracy for the same. With the comparative analysis, it is observed that the testing and training accuracy is 83.47% to 86.80%, respectively by using only 62.5% features. Hypothetically it is proved that there is no variance between the generalization performance of F-ELM (with all features or before feature selection) and RFELM++ (after feature selection) for four samples. A novel application, IoT-based precision agriculture expert system, is developed by exploiting the proposed algorithms RFELM++ and RFOSELM++. Currently, the algorithms are used for binary classification in future it can be used for multi-class classification. The precision system can be improved by considering the rain removal [73] and by using CNN [74].

Fig 1.

Figure 1.

Extreme learning machine.

The International Journal of Fuzzy Logic and Intelligent Systems 2023; 23: 465-481https://doi.org/10.5391/IJFIS.2023.23.4.465

Fig 2.

Figure 2.

RELM++ and RFELM++ by using relief algorithm.

The International Journal of Fuzzy Logic and Intelligent Systems 2023; 23: 465-481https://doi.org/10.5391/IJFIS.2023.23.4.465

Fig 3.

Figure 3.

Comparison of ELM and F-ELM: (a) accuracy of PID dataset, (b) accuracy of SHD dataset, (c) precision of PID dataset, (d) precision of SHD dataset, (e) recall of PID dataset, and (f) recall of SHD dataset.

The International Journal of Fuzzy Logic and Intelligent Systems 2023; 23: 465-481https://doi.org/10.5391/IJFIS.2023.23.4.465

Fig 4.

Figure 4.

Accuracy measure comparison of F-ELM with various classifiers by using Relief-FSS algorithm: (a) input PID dataset and (b) input SHD dataset.

The International Journal of Fuzzy Logic and Intelligent Systems 2023; 23: 465-481https://doi.org/10.5391/IJFIS.2023.23.4.465

Fig 5.

Figure 5.

Evaluation performance: (a) CHD accuracy, (b) CHD precision, and (c) CHD recall.

The International Journal of Fuzzy Logic and Intelligent Systems 2023; 23: 465-481https://doi.org/10.5391/IJFIS.2023.23.4.465

Fig 6.

Figure 6.

ELM and F-ELM performance comparison with incremental learning for training.

The International Journal of Fuzzy Logic and Intelligent Systems 2023; 23: 465-481https://doi.org/10.5391/IJFIS.2023.23.4.465

Fig 7.

Figure 7.

ELM and F-ELM performance comparison with incremental learning for testing.

The International Journal of Fuzzy Logic and Intelligent Systems 2023; 23: 465-481https://doi.org/10.5391/IJFIS.2023.23.4.465

Fig 8.

Figure 8.

Performance metrics of RFELM++ before and after using Relief-FSS algorithm.

The International Journal of Fuzzy Logic and Intelligent Systems 2023; 23: 465-481https://doi.org/10.5391/IJFIS.2023.23.4.465

Fig 9.

Figure 9.

F-ELM and ELM in combination with different FSS methods for PID dataset.

The International Journal of Fuzzy Logic and Intelligent Systems 2023; 23: 465-481https://doi.org/10.5391/IJFIS.2023.23.4.465

Fig 10.

Figure 10.

Conceptual framework for the IoT-based precision agriculture expert system.

The International Journal of Fuzzy Logic and Intelligent Systems 2023; 23: 465-481https://doi.org/10.5391/IJFIS.2023.23.4.465

Fig 11.

Figure 11.

ELM and RELM++ (testing accuracy) for plant disease and nutrient deficiency.

The International Journal of Fuzzy Logic and Intelligent Systems 2023; 23: 465-481https://doi.org/10.5391/IJFIS.2023.23.4.465

Table 1 . Comparative results of ELM and F-ELM with different FS methods for input PID dataset.

FSS MethodsELMF-ELMFeature subset
MS78.2679.132, 6, 8
HS; l76.0882.61, 2, 6, 8
NN for threshold selection78.2682.61, 2, 6, 8
PSO78.2682.61, 2, 6, 8
SRLPSO78.2679.132, 6, 8
t-test76.5283.471, 2, 6, 7, 8
Without FSS Method71.7380.43All
Avg76.7681.42-

Table 2 . Parameters for RFELM++.

ParameterFeatures
Input neuronSame as input features
Output neuron2
Hidden neuron10:5:200
Activity functionSigmoidal
Sequential mode1 by 1, 20 by 20, 10 by 30
Dataset divisionTraining (70%)-Testing (30%)

Table 3 . Comparison of RFELM++ with existing learning algorithms for input PID dataset.

Methods/algorithmAccuracy of trainingAccuracy of testingTotal number of attributes used
RFELM++86.8083.4705 (62.5%)
F-ELM [59]75.3574.0908 (100%)
MFMM-FIS [60]n/a72.9208 (100%)
EGART-FIS [61]n/a73.0508 (100%)
MFMM-FIS [62]n/a72.9208 (100%)

Table 4 . RFELM++ and existing learning algorithms comparison for input PID dataset.

Method/algorithmAccuracy (testing)Total number of attributes usedFeature subset
RFELM++83.4705{1,2,6,7,8}
RELM++76.5205{1,2,6,7,8}
EGART-FIS [61]73.0508{1,2,3,4,5,6,7,8}
F-ELM [59]74.0908{1,2,3,4,5,6,7,8}
MFMM-FIS [60]72.9208{1,2,3,4,5,6,7,8}
LS-ELM [65]78.2104{1,2,6,8}
GDA-LS-ELM [65]79.1605{1,2,6,7,8}
MLNN with LM [64]79.6204{1,2,6,8}
PNN [64]78.0503{2,6,8}
MS [62]76.0403{2,6,8}
HS [62]75.9104{1,2,6,8}
NN [62]76.0403{2,6,8}
GP-KNN [66]80.5008{1,2,3,4,5,6,7,8}
Gini-Fuzzy [67]75.8008{1,2,3,4,5,6,7,8}
FMNN-CART-RF [25]78.3908{1,2,3,4,5,6,7,8}
Generic statistical approach [68]78.0408{1,2,3,4,5,6,7,8}
Average76.8426.79-

Table 5 . RFELM++ and existing learning algorithms comparison for Input SHD dataset.

Method/algorithmAccuracy (testing)Number of featuresFeature subset
RFELM++92.5906{3,9,10,11,12,13}
RELM++90.7405{3,9,10,12,13}
ESUNN [69]83.2205{3,8,9,11,12}
EPUNN [69]81.8904{8,9,11,12}
MR+EPUNN [70]83.1205{8,9,11,12,13}
MS [62]84.4406{3,8,9,11,12,13}
HS [62]84.8107{3,8,9,10,11,12,13}
NN [62]85.1904{3,11,12,13}
Average83.7760.26-

Table 6 . Evaluation comparative analysis for ELM and F-ELM with different FSS for input PID dataset.

FSS MethodELMF-ELMFeature subset
LS [65]78.2182.6{1,2,6,8}
GDA-LS [65]79.1682.17{1,2,6,7,8}
MLNN with LM [64]79.6282.6{1,2,6,8}
PNN [64]78.0579.13{2,6,8}
LDA-MW [62]89.7482.17{1,2,6,7,8}
MS [62]76.0479.13{2,6,8}
HS [62]75.9182.6{1,2,6,8}
NN [62]76.0479.13{2,6,8}

References

  1. Zhang, WB, and Ji, HB (2013). Fuzzy extreme learning machine for classification. Electronics Letters. 49, 448-450. https://doi.org/10.1049/el.2012.3642
    CrossRef
  2. Krawczyk, B, Wozniak, M, and Herrera, F . Weighted one-class classification for different types of minority class examples in imbalanced data., Proceedings of IEEE Symposium on Computational Intelligence and Data Mining (CIDM), 2014, Orlando, FL, USA, Array, pp.337-344. https://doi.org/10.1109/CIDM.2014.7008687
    CrossRef
  3. Juan, J, and Cubero, C . Weighted classification using decision trees for binary classification problems., Proceedings of the Spanish Computer Science Congress (CEDI 2007, II Congreso Español de Informática), 2007, Zaragoza, Spain.
  4. Eitrich, T, and Lang, B . On the advantages of weighted l1-norm support vector learning for unbalanced binary classification problems., Proceedings of 2006 3rd International IEEE Conference on Intelligent Systems, 2006, London, UK, Array, pp.575-580. https://doi.org/10.1109/IS.2006.348483
    CrossRef
  5. Hwang, JP, Park, S, and Kim, E (2011). A new weighted approach to imbalanced data classification problem via support vector machine with quadratic cost function. Expert Systems with Applications. 38, 8580-8585. https://doi.org/10.1016/j.eswa.2011.01.061
    CrossRef
  6. Lee, CH (2015). A gradient approach for value weighted classification learning in naive bayes. Knowledge-Based Systems. 85, 71-79. https://doi.org/10.1016/j.knosys.2015.04.020
    CrossRef
  7. Kale, AP, and Sonavane, S (2018). PF-FELM: a robust PCA feature selection for fuzzy extreme learning machine. IEEE Journal of Selected Topics in Signal Processing. 12, 1303-1312. https://doi.org/10.1109/JSTSP.2018.2873988
    CrossRef
  8. Yu, L, and Liu, H (2004). Efficient feature selection via analysis of relevance and redundancy. Journal of Machine Learning Research. 5, 1205-1224.
  9. Dash, M, and Liu, H (1997). Feature selection for classification. Intelligent Data Analysis. 1, 131-156. https://doi.org/10.1016/S1088-467X(97)00008-5
    CrossRef
  10. Forman, G (2003). An extensive empirical study of feature selection metrics for text classification. Journal of Machine Learning Research. 3, 1289-1305.
  11. Bharathi, P, and Subashini, P . Optimal feature subset selection using differential evolution with sequential extreme learning machine for river ice images., Proceedings of IEEE Region 10 Conference (TENCON), 2015, Macao, China, Array, pp.1-6. https://doi.org/10.1109/TENCON.2015.7372861
    CrossRef
  12. de Assis Boldt, F, Rauber, TW, Varejao, FM, and Ribeiro, MP . Fast feature selection using hybrid ranking and wrapper approach for automatic fault diagnosis of motorpumps based on vibration signals., Proceedings of 2015 IEEE 13th International Conference on Industrial Informatics (INDIN), 2015, Cambridge, UK, Array, pp.127-132. https://doi.org/10.1109/INDIN.2015.7281722
    CrossRef
  13. Tiwari, S, and Singh, B . A hybrid approach for feature selection., Proceedings of 2015 3rd International Conference on Image Information Processing (ICIIP), 2015, Waknaghat, India, Array, pp.277-280. https://doi.org/10.1109/ICIIP.2015.7414780
    CrossRef
  14. Das, S . Filters, wrappers and a boosting-based hybrid for feature selection., Proceedings of the Eighteenth International Conference on Machine Learning (ICML), 2001, Williamstown, MA, USA, pp.74-81.
  15. Tahir, GA, and Loo, CK (2020). An open-ended continual learning for food recognition using class incremental extreme learning machines. IEEE Access. 8, 82328-82346. https://doi.org/10.1109/ACCESS.2020.2991810
    CrossRef
  16. Zhang, T, Ding, B, Zhao, X, and Yue, Q (2018). A fast feature selection algorithm based on swarm intelligence in acoustic defect detection. IEEE Access. 6, 28848-28858. https://doi.org/10.1109/ACCESS.2018.2833164
    CrossRef
  17. bin Mohd Zain, MZ, Kanesan, J, Chuah, JH, Dhanapal, S, and Kendall, G (2018). A multi-objective particle swarm optimization algorithm based on dynamic boundary search for constrained optimization. Applied Soft Computing. 70, 680-700. https://doi.org/10.1016/j.asoc.2018.06.022
    CrossRef
  18. Zhang, GP (2000). Neural networks for classification: a survey. IEEE Transactions on Systems, Man, and Cybernetics Part C (Applications and Reviews). 30, 451-462. https://doi.org/10.1109/5326.897072
    CrossRef
  19. Tsipouras, MG, Voglis, C, and Fotiadis, DI (2007). A framework for fuzzy expert system creation: application to cardiovascular diseases. IEEE Transactions on Biomedical Engineering. 54, 2089-2105. https://doi.org/10.1109/TBME.2007.893500
    Pubmed CrossRef
  20. Tao, X, Zhang, D, Wang, Z, Liu, X, Zhang, H, and Xu, D (2020). Detection of power line insulator defects using aerial images analyzed with convolutional neural networks. IEEE Transactions on Systems, Man, and Cybernetics: Systems. 50, 1486-1498. https://doi.org/10.1109/TSMC.2018.2871750
    CrossRef
  21. Wen, L, Gao, L, and Li, X (2019). A new deep transfer learning based on sparse auto-encoder for fault diagnosis. IEEE Transactions on Systems, Man, and Cybernetics: Systems. 49, 136-144. https://doi.org/10.1109/TSMC.2017.2754287
    CrossRef
  22. Mojrian, S, Pinter, G, Joloudari, JH, Felde, I, Szabo-Gali, A, Nadai, L, and Mosavi, A . Hybrid machine learning model of extreme learning machine radial basis function for breast cancer detection and diagnosis; a multilayer fuzzy expert system., Proceedings of 2020 RIVF International Conference on Computing and Communication Technologies (RIVF), 2020, Ho Chi Minh City, Vietnam, Array, pp.1-7. https://doi.org/10.1109/RIVF48685.2020.9140744
    CrossRef
  23. Zhang, Y, Wu, J, Cai, Z, Zhang, P, and Chen, L (2016). Memetic extreme learning machine. Pattern Recognition. 58, 135-148. https://doi.org/10.1016/j.patcog.2016.04.003
    CrossRef
  24. Huang, GB, Zhu, QY, and Siew, CK (2006). Extreme learning machine: theory and applications. Neurocomputing. 70, 489-501. https://doi.org/10.1016/j.neucom.2005.12.126
    CrossRef
  25. Seera, M, and Lim, CP (2014). A hybrid intelligent system for medical data classification extreme learning machine: theory and applications. Expert System Application. 41, 2239-2249. https://doi.org/10.1016/j.eswa.2013.09.022
    CrossRef
  26. Karakaya, G, Galelli, S, Ahipasaoglu, SD, and Taormina, R (2016). Identifying (quasi) equally informative subsets in feature selection problems for classification: a max-relevance min-redundancy approach. IEEE Transactions on Cybernetics. 46, 1424-1437. https://doi.org/10.1109/TCYB.2015.2444435
    Pubmed CrossRef
  27. Huynh, HT, Won, Y, and Kim, J (2015). Hematocrit estimation using online sequential extreme learning machine. Biomedical Materials and Engineering. 26, S2025-S2032. https://doi.org/10.3233/BME-151507
    Pubmed CrossRef
  28. Singh, RP, Dabas, N, Chaudhary, V, and Nagendra, (2016). Online sequential extreme learning machine for watermarking in DWT domain. Neurocomputing. 174, 238-249. https://doi.org/10.1016/j.neucom.2015.03.115
    CrossRef
  29. Singh, R, Kumar, H, and Singla, R (2015). An intrusion detection system using network traffic profiling and online sequential extreme learning machine. Expert Systems with Applications. 42, 8609-8624. https://doi.org/10.1016/j.eswa.2015.07.015
    CrossRef
  30. Rong, HJ, Huang, GB, Sundararajan, N, and Saratchandran, P (2009). Online sequential fuzzy extreme learning machine for function approximation and classification problems. IEEE Transactions on Systems, Man, and Cybernetics Part B (Cybernetics). 39, 1067-1072. https://doi.org/10.1109/TSMCB.2008.2010506
    Pubmed CrossRef
  31. Ontiveros-Robles, E, Castillo, O, and Melin, P (2020). An approach for non-singleton generalized type-2 fuzzy classifiers. Journal of Intelligent & Fuzzy Systems. 35, 7203-7215. https://doi.org/10.3233/JIFS-200639
    CrossRef
  32. Ontiveros, E, Melin, P, and Castillo, O (2020). Designing hybrid classifiers based on general type-2 fuzzy logic and support vector machines. Soft Computing. 24, 18009-18019. https://doi.org/10.1007/s00500-020-05052-x
    CrossRef
  33. Rubio, E, Castillo, O, Valdez, F, Melin, P, Gonzalez, CI, and Martinez, G (2017). An extension of the fuzzy possibilistic clustering algorithm using type-2 fuzzy logic techniques. Advances in Fuzzy Systems. 2017. article no 7094046. https://doi.org/10.1155/2017/7094046
    CrossRef
  34. Huang, GB, Chen, L, and Siew, CK (2006). Universal approximation using incremental constructive feedforward networks with random hidden nodes. IEEE Transactions on Neural Networks. 17, 879-892.
    Pubmed CrossRef
  35. Huang, GB, and Chen, L (2007). Convex incremental extreme learning machine. Neurocomputing. 70, 3056-3062. https://doi.org/10.1016/j.neucom.2007.02.009
    CrossRef
  36. Gepperth, A, and Hammer, B . Incremental learning algorithms and applications., Proceedings of European Symposium on Artificial Neural Networks (ESANN), 2016, Bruges, Belgium.
  37. Bermejo, P, de La Ossa, L, Gamez, JA, and Puerta, JM (2011). Enhancing incremental feature subset selection in high-dimensional databases by adding a backward step. Computer and Information Sciences II. London, UK: Springer, pp. 93-97 https://doi.org/10.1007/978-1-4471-2155-8_11
    CrossRef
  38. Perkins, S, Lacker, K, and Theiler, J (2003). Grafting: fast, incremental feature selection by gradient descent in function space. Journal of Machine Learning Research. 3, 1333-1356.
  39. Bermejo, P, Gamez, J, and Puerta, J . On incremental wrapper-based attribute selection: experimental analysis of the relevance criteria., Proceedings of the 12th International Conference Information Processing and Management of Uncertainty for Knowledge-Based Systems (IPMU), 2008, Torremolinos, Malaga, Spain, pp.638-645.
  40. Ruiz, R, Riquelme, JC, and Aguilar-Ruiz, JS (2006). Incremental wrapper-based gene selection from microarray data for cancer classification. Pattern Recognition. 39, 2383-2392. https://doi.org/10.1016/j.patcog.2005.11.001
    CrossRef
  41. d’Alche-Buc, F, and Ralaivola, L (2002). Incremental learning algorithms for classification and regression: Local strategies. AIP Conference Proceedings. 627, 320-329. https://doi.org/10.1063/1.1503700
    CrossRef
  42. Feng, G, Huang, GB, Lin, Q, and Gay, R (2009). Error minimized extreme learning machine with growth of hidden nodes and incremental learning. IEEE Transactions on Neural Networks. 20, 1352-1357. https://doi.org/10.1109/TNN.2009.2024147
    Pubmed CrossRef
  43. Yang, Y, Wang, Y, and Yuan, X (2012). Bidirectional extreme learning machine for regression problem and its learning effectiveness. IEEE Transactions on Neural Networks and Learning Systems. 23, 1498-1505. https://doi.org/10.1109/TNNLS.2012.2202289
    Pubmed CrossRef
  44. Baluja, S (1994). Population-based incremental learning: a method for integrating genetic search based function optimization and competitive learning. Pittsburgh, PA, USA: Carnegie Mellon University Technical Report No. CMU-CS-94-163
  45. Wang, A, An, N, Chen, G, Yang, J, Li, L, and Alterovitz, G . Incremental wrapper based gene selection with Markov blanket., Proceedings of 2014 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), 2014, Belfast, UK, Array, pp.74-79. https://doi.org/10.1109/BIBM.2014.6999251
    CrossRef
  46. Kale, A, and Sonavane, S (2019). F-WSS++: incremental wrapper subset selection algorithm for fuzzy extreme learning machine. International Journal of Machine Learning and Cybernetics. 10, 1821-1832. https://doi.org/10.1007/s13042-018-0859-9
    CrossRef
  47. Wang, S, Ouyang, L, Yuan, Y, Ni, X, Han, X, and Wang, FY (2019). Blockchain-enabled smart contracts: architecture, applications, and future trends. IEEE Transactions on Systems, Man, and Cybernetics: Systems. 49, 2266-2277. https://doi.org/10.1109/TSMC.2019.2895123
    CrossRef
  48. UC Irvine Machine Learning Repository. [online] Available: http://archive.ics.uci.edu/
  49. Sun, Y, and Li, J . Iterative relief for feature weighting., Proceedings of the 23rd International Conference on Machine Learning, 2006, Pittsburgh, PA, USA, Array, pp.913-920. https://doi.org/10.1145/1143844.1143959
    Pubmed CrossRef
  50. Hall, MA, and Holmes, G (2003). Benchmarking attribute selection techniques for discrete class data mining. IEEE Transactions on Knowledge and Data Engineering. 15, 1437-1447. https://doi.org/10.1109/TKDE.2003.1245283
    CrossRef
  51. Urbanowicz, RJ, Meeker, M, La Cava, W, Olson, RS, and Moore, JH (2018). Relief-based feature selection: introduction and review. Journal of Biomedical Informatics. 85, 189-203. https://doi.org/10.1016/j.jbi.2018.07.014
    Pubmed KoreaMed CrossRef
  52. Nahato, KB, Nehemiah, KH, and Kannan, A (2016). Hybrid approach using fuzzy sets and extreme learning machine for classifying clinical datasets. Informatics in Medicine Unlocked. 2, 1-11. https://doi.org/10.1016/j.imu.2016.01.001
    CrossRef
  53. Mahdiyah, U, Irawan, MI, and Imah, EM (2015). Integrating data selection and extreme learning machine for imbalanced data. Procedia Computer Science. 59, 221-229. https://doi.org/10.1016/j.procs.2015.07.561
    CrossRef
  54. Subbulakshmi, C, and Deepa, S (2015). Medical dataset classification: a machine learning paradigm integrating particle swarm optimization with extreme learning machine classifier. The Scientific World Journal. 2015. article no 418060. https://doi.org/10.1155/2015/418060
    Pubmed KoreaMed CrossRef
  55. Novakovic, J, Strbac, P, and Bulatovic, D (2011). Toward optimal feature selection using ranking methods and classification algorithms. Yugoslav Journal of Operations Research. 21, 119-135. https://doi.org/10.2298/YJOR1101119N
    CrossRef
  56. Wuensch, KL (1996). Straightforward statistics for the behavioral sciences. Journal of the American Statistical Association. 91, 1750-1751. https://doi.org/10.2307/2291607
    CrossRef
  57. Devaraj, S, and Paulraj, S (2015). An efficient feature subset selection algorithm for classification of multidimensional dataset. The Scientific World Journal. 2015. article no 821798. https://doi.org/10.1155/2015/821798
    Pubmed KoreaMed CrossRef
  58. Kothari, CR (2004). Research Methodology: Methods and Techniques. New Delhi, India: New Age International
  59. Wong, SY, Yap, KS, Yap, HJ, Tan, SC, and Chang, SW (2015). On equivalence of FIS and ELM for interpretable rule-based knowledge representation. IEEE Transactions on Neural Networks and Learning Systems. 26, 1417-1430. https://doi.org/10.1109/TNNLS.2014.2341655
    Pubmed CrossRef
  60. Quteishat, A, and Lim, CP (2008). A modified fuzzy min-max neural network with rule extraction and its application to fault detection and classification. Applied Soft Computing. 8, 985-995. https://doi.org/10.1016/j.asoc.2007.07.013
    CrossRef
  61. Yap, KS, Lim, CP, and Mohamad-Saleh, J (2010). An enhanced generalized adaptive resonance theory neural network and its application to medical pattern classification. Journal of Intelligent & Fuzzy Systems. 21, 65-78. https://doi.org/10.3233/IFS-2010-0436
    CrossRef
  62. Calisir, D, and Dogantekin, E (2011). An automatic diabetes diagnosis system based on lda-wavelet support vector machine classifier. Expert Systems with Applications. 38, 8311-8315. https://doi.org/10.1016/j.eswa.2011.01.017
    CrossRef
  63. Jaganathan, P, and Kuppuchamy, R (2013). A threshold fuzzy entropy based feature selection for medical database classification. Computers in Biology and Medicine. 43, 2222-2229. https://doi.org/10.1016/j.compbiomed.2013.10.016
    Pubmed CrossRef
  64. Temurtas, H, Yumusak, N, and Temurtas, F (2009). A comparative study on diabetes disease diagnosis using neural networks. Expert Systems with Applications. 36, 8610-8615. https://doi.org/10.1016/j.eswa.2008.10.032
    CrossRef
  65. Polat, K, Gunes, S, and Arslan, A (2008). A cascade learning system for classification of diabetes disease: generalized discriminant analysis and least square support vector machine. Expert Systems with Applications. 34, 482-487. https://doi.org/10.1016/j.eswa.2006.09.012
    CrossRef
  66. Aslam, MW, Zhu, Z, and Nandi, AK (2013). Feature generation using genetic programming with comparative partner selection for diabetes classification. Expert Systems with Applications. 40, 5402-5412. https://doi.org/10.1016/j.eswa.2013.04.003
    CrossRef
  67. Varma, KV, Rao, AA, Lakshmi, TSM, and Rao, PN (2014). A computational intelligence approach for a better diagnosis of diabetic patients. Computers & Electrical Engineering. 40, 1758-1765. https://doi.org/10.1016/j.compeleceng.2013.07.003
    CrossRef
  68. Mattila, J, Koikkalainen, J, Virkki, A, Van Gils, M, and Lotjonen, J (2012). Design and application of a generic clinical decision support system for multiscale data. IEEE Transactions on Biomedical Engineering. 59, 234-240. https://doi.org/10.1109/TBME.2011.2170986
    Pubmed KoreaMed CrossRef
  69. Martinez-Estudillo, FJ, Hervas-Martinez, C, Gutierrez, PA, and Martinez-Estudillo, AC (2008). Evolutionary product-unit neural networks classifiers. Neurocomputing. 72, 548-561. https://doi.org/10.1016/j.neucom.2007.11.019
    CrossRef
  70. Hervas-Martinez, C, Martinez-Estudillo, FJ, and Carbonero-Ruz, M (2008). Multilogistic regression by means of evolutionary product-unit neural networks. Neural Networks. 21, 951-961. https://doi.org/10.1016/j.neunet.2007.12.052
    Pubmed CrossRef
  71. Zhang, X, and Zhang, F . Images features extraction of tobacco leaves., Proceedings of 2008 Congress on Image and Signal Processing, 2008, Sanya, China, Array, pp.773-776. https://doi.org/10.1109/CISP.2008.88
    CrossRef
  72. Ismail, N, Othman, ZA, and Samsudin, NA (2019). Regularization activation function for extreme learning machine. International Journal of Advanced Computer Science and Applications. 10, 241-248. https://doi.org/10.14569/ijacsa.2019.0100331
    CrossRef
  73. Mi, Y, Yuan, S, Li, X, and Zhou, J (2021). Dense residual generative adversarial network for rapid rain removal. IEEE Access. 9, 24848-24858. https://doi.org/10.1109/ACCESS.2021.3055527
    CrossRef
  74. Garcia-Gasulla, D, Pares, F, Vilalta, A, Moreno, J, Ayguadé, E, Labarta, J, Cortes, U, and Suzumura, T (2018). On the behavior of convolutional nets for feature extraction. Journal of Artificial Intelligence Research. 61, 563-592. https://doi.org/10.1613/jair.5756
    CrossRef

Share this article on :

Related articles in IJFIS