A data-driven machine learning approach toward an improved maize crop production

Abstract

The increasing need for improved approaches to sustainable food production to meet the ever-growing global population, especially in sub-Saharan Africa cannot be undermined. The agricultural industry has attracted a series of technological advancements towards improved food production, preservation, and sustainable farm practices. The technologies that are playing significant roles include Machine Learning (ML), Artificial Intelligence (AI), Internet of Things (IoT) among others. This study uses a dataset collected in a sub-Saharan African farm to model the maize crop yield improvement. The dataset combines soil parameters, and atmospheric parameters as well as the physical parameters of the maize plants over their life span. This work explores the performance of six (6) unique machine learning models namely: Nave Bayes (NB), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Decision Trees (DT), Artificial Neural Networks (ANN), and hybridized ANN-KNN. In the performance evaluation of these classifiers, the hybridized ANN-KNN shows an accuracy of 99.45% building the performance of the ANN classifiers with an accuracy of 98.83%. The performances of other classifiers are 96.04%, 92.49%, 78.78%, and the lowest accuracy of 53.88% for DT, KNN, SVM, and NB respectively. The result underscores the importance of hybridization in improving machine learning performance for predicting maize crop yield. The improved accuracy achieved can serve as invaluable resources for farmers in making informed decisions regarding crop selection, soil treatment, plant treatments, and cultivation strategies.