IMPLEMENTATION OF MACHINE LEARNING FOR PREDICTING MAIZE CROP YIELDS USING MULTIPLE LINEAR REGRESSION AND BACKWARD ELIMINATION
Stephen Gbenga Fashoto1*,Elliot Mbunge2, Gabriel Ogunleye3, and Johan Van den Burg4
1Department of Computer Science, University of Eswatini, Kwaluseni, Eswatini (formerly Swaziland),
2Department of Computer Science, Federal University, Oye-Ekiti, Nigeria
1*This email address is being protected from spambots. You need JavaScript enabled to view it., 2This email address is being protected from spambots. You need JavaScript enabled to view it., 3This email address is being protected from spambots. You need JavaScript enabled to view it., 4This email address is being protected from spambots. You need JavaScript enabled to view it.
ABSTRACT
Predicting maize crop yields especially in maize production is paramount in order to alleviate poverty and contribute towards food security. Many regions experience food shortage especially in Africa because of uncertain climatic changes, poor irrigation facilities, reduction in soil fertility and traditional farming techniques. Therefore, predicting maize crop yields helps policymakers to make timely import and export decisions to strengthen national food security. However, none of the published work has been done to predict maize crop yields using machine learning in Eswatini, Africa. This paper aimed at applying machine learning (ML) to predict maize yields for a single season in Eswatini. A ML model was trained and tested using open-source data and local data. This is done by using three different data splits with the opensource predictor data consisting of 48 data points each with 7 attributes and open-source response data consisting of 48 data points each with a single attribute, adjusted R² values were 0.784 (at 70:30), 0.849 (at 80:20), and 0.878 (at 90:10) before being normalized, 1.00 across the board after normalization, and 0.846 (at 70:30), 0.886 (at 80:20), and 0.885 (at 90:10) after backward elimination. At the second attempt, it is done by using the combined predictor data of 68 data points with 7 attributes each and combined response data of 68 data points with a single attribute each, with the same data splits and methods adjusted R² values were 0.966 (at 70:30), 0.972 (at 80:20), and 0.978 (at 90:10) before being normalized, 1.00 across the board after normalization, and 0.967 (at 70:30), 0.973 (at 80:20), and 0.978 (at 90:10) after backward elimination.
Keywords: Agricultural technology, Backward elimination, Environmental factors, Linear Regression Machine Learning, Maize crop.
Published On: 29 January 2021
