International Journal of Fuzzy Logic and Intelligent Systems 2023; 23(2): 214-228
Published online June 25, 2023
https://doi.org/10.5391/IJFIS.2023.23.2.214
© The Korean Institute of Intelligent Systems
Christine Musanase1,2, Anthony Vodacek2,3 , Damien Hanyurwimfura1,2 , Alfred Uwitonze1,2 , Aloys Fashaho1,2 , and Adrien Turamyemyirijuru1,2
1African Center of Excellence in Internet of Things, University of Rwanda, Kigali, Rwanda
2Chester F. Carlson Center for Imaging Science, Rochester Institute of Technology, Rochester, NY, USA
3College of Agriculture, Animal Sciences and Veterinary Medicine, University of Rwanda, Musanze, Rwanda
Correspondence to :
Christine Musanase (musanasechristine@gmail.com)
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.
The ability to estimate soil quality has great value for agriculture, especially for low-income regions with minimal agricultural and financial resources. This prediction provides users with information that is useful in determining whether the soil is suitable for a specific crop, such as potato (Solanum tuberosum). Farmers in Rwanda lack information on soil quality. There are not enough soil laboratories to perform the requisite measurements of NPK, pH, and organic carbon, nor are there enough experts to analyze the data and provide farmers with timely results. The prime objective of the proposed study is to develop a predictive framework that can estimate soil quality for the ideal cultivation of potato (Solanum tuberosum) considering a case study of Rwanda. In this study, bootstrapping is used to augment the small soil dataset, and fuzzy logic is used to label soil data into four classes of soil suitability, with verification of the labeling by soil experts. Several machine learning methods are then tested on the labeled data, resulting in the classification of suitability for the augmented dataset and an assessment of their performance as a way to support experts in predicting soil quality. All machine learning methods applied were viable, with the best performance achieved using an artificial neural network. The quantified outcome showed that the adoption of a neural-network-based scheme has an average accuracy of 32% in contrast to other learning schemes. However, 70%-80% accuracy was achieved upon the adoption of fuzzy logic.
Keywords: Soil quality, Fuzzy logic, Artificial intelligence, Rwanda, Machine learning, NPK, Predictive model
The authors declare no conflicts of interest.
International Journal of Fuzzy Logic and Intelligent Systems 2023; 23(2): 214-228
Published online June 25, 2023 https://doi.org/10.5391/IJFIS.2023.23.2.214
Copyright © The Korean Institute of Intelligent Systems.
Christine Musanase1,2, Anthony Vodacek2,3 , Damien Hanyurwimfura1,2 , Alfred Uwitonze1,2 , Aloys Fashaho1,2 , and Adrien Turamyemyirijuru1,2
1African Center of Excellence in Internet of Things, University of Rwanda, Kigali, Rwanda
2Chester F. Carlson Center for Imaging Science, Rochester Institute of Technology, Rochester, NY, USA
3College of Agriculture, Animal Sciences and Veterinary Medicine, University of Rwanda, Musanze, Rwanda
Correspondence to:Christine Musanase (musanasechristine@gmail.com)
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.
The ability to estimate soil quality has great value for agriculture, especially for low-income regions with minimal agricultural and financial resources. This prediction provides users with information that is useful in determining whether the soil is suitable for a specific crop, such as potato (Solanum tuberosum). Farmers in Rwanda lack information on soil quality. There are not enough soil laboratories to perform the requisite measurements of NPK, pH, and organic carbon, nor are there enough experts to analyze the data and provide farmers with timely results. The prime objective of the proposed study is to develop a predictive framework that can estimate soil quality for the ideal cultivation of potato (Solanum tuberosum) considering a case study of Rwanda. In this study, bootstrapping is used to augment the small soil dataset, and fuzzy logic is used to label soil data into four classes of soil suitability, with verification of the labeling by soil experts. Several machine learning methods are then tested on the labeled data, resulting in the classification of suitability for the augmented dataset and an assessment of their performance as a way to support experts in predicting soil quality. All machine learning methods applied were viable, with the best performance achieved using an artificial neural network. The quantified outcome showed that the adoption of a neural-network-based scheme has an average accuracy of 32% in contrast to other learning schemes. However, 70%-80% accuracy was achieved upon the adoption of fuzzy logic.
Keywords: Soil quality, Fuzzy logic, Artificial intelligence, Rwanda, Machine learning, NPK, Predictive model
Structure of the proposed scheme of implementation.
Study area from which the soil data were derived, i.e., Rubavu, Burera, Gicumbi, and Rwamagana.
Architecture of fuzzy logic type 2.
Histogram of OC percent.
N percentage histogram.
P percentage histogram.
K percentage histogram.
Water pH histogram.
Results of the fuzzy logic labeling for the three different samples. The classes are ordered by prevalence.
Comparison of all algorithms.
Algorithm 1. Algorithm for labeling using fuzzy logic type-2..
1. | Add 4 new columns in S |
-c1 for pH quality | |
-c2 for N quality, | |
-c3 for P quality | |
-c4 for K quality | |
2. | Assign quality values according to the study conducted |
3. | If pH quality is 1 |
-if N quality is 1 then quality is 1 | |
-else quality is smallest among p & k | |
4. | Else if pH quality is 2 |
-if n quality is 1 then quality is 1 | |
-else quality is 2 | |
5. | Quality is equal to pH quality |
Table 1. Dataset sample presentation of values obtained from soil analysis.
pH | OC (%) | N (%) | P (ppm) | K (ppm) |
---|---|---|---|---|
5 | 2.88 | 0.09 | 18.8 | 80.3 |
4.89 | 2.93 | 0.07 | 29.3 | 73.5 |
4.94 | 2.91 | 0.07 | 8.37 | 45.1 |
5.3 | 2.78 | 0.10 | 18.8 | 88 |
5.0 | 2.71 | 0.07 | 13.6 | 51 |
4.94 | 2.65 | 0.077 | 8.37 | 87 |
Table 2. Classification of selected soil properties values for potato.
Suitable | Moderately suitable | Marginally suitable | Not suitable | |
---|---|---|---|---|
K (ppm) | >55 | 35–55 | 15–35 | <15 |
P (ppm) | >10 | 6.5–10 | 2.5–6.5 | <2.5 |
N (%) | >0.30 | 0.225–0.30 | 0.125–0.225 | <0.125 |
OC (%) | >0.7 | 0.5–0.7 | 0.3–0.5 | <0.3 |
pH | 5.5–7 | 5–5.5 | 4–5 | <4 |
7–7.5 | 7.5–8 | >8 |
Table 3. Numerical outcome of accuracy-based comparative analysis.
ML approaches | Precision (%) | Recall (%) | F1-score |
---|---|---|---|
Gaussian NB | 0.82 | 0.83 | 0.84 |
ANN | 0.87 | 0.85 | 0.86 |
Logistic regression | 0.79 | 0.75 | 0.77 |
KNN | 0.75 | 0.72 | 0.73 |
LeeChae Jang,TaeKyun Kim,Dal-Won Park,Daniela A. Langova-Orozova
Int. J. Fuzzy Log. Intell. Syst. 2004; 4(1): 90-96Amirthalakshmi Thirumalai Maadapoosi, Velan Balamurugan, V. Vedanarayanan, Sahaya Anselin Nisha, and R. Narmadha
International Journal of Fuzzy Logic and Intelligent Systems 2024; 24(3): 231-241 https://doi.org/10.5391/IJFIS.2024.24.3.231Abdul Kareem and Varuna Kumara
International Journal of Fuzzy Logic and Intelligent Systems 2024; 24(1): 30-42 https://doi.org/10.5391/IJFIS.2024.24.1.30Structure of the proposed scheme of implementation.
|@|~(^,^)~|@|Study area from which the soil data were derived, i.e., Rubavu, Burera, Gicumbi, and Rwamagana.
|@|~(^,^)~|@|Architecture of fuzzy logic type 2.
|@|~(^,^)~|@|Histogram of OC percent.
|@|~(^,^)~|@|N percentage histogram.
|@|~(^,^)~|@|P percentage histogram.
|@|~(^,^)~|@|K percentage histogram.
|@|~(^,^)~|@|Water pH histogram.
|@|~(^,^)~|@|Results of the fuzzy logic labeling for the three different samples. The classes are ordered by prevalence.
|@|~(^,^)~|@|Comparison of all algorithms.