A Machine Learning Approach to Season Prediction Based on Agricultural Input Orders

Abstract

Agriculture is a source of income to many people around the globe. The evolution of modern agriculture led to an increase in the use of modern agricultural inputs, which has led to a sustainable growth in crop production. Based on their previous experiences, agricultural input suppliers stock a lot of products in their stores as they wait for right seasons. Unfortunately, due to the lack of precise knowledge about seasonal changes, their products overstay and some get tainted, contaminated, and expired leading to losses caused by dumping because of low sales made; hence, the closure of some of the businesses. In this fourth industrial revolution, which is smart agriculture based, agricultural input suppliers need the help of artificial intelligence to facilitate decision-making that can make their businesses more profitable and sustainable. In this research, we help suppliers predict the current season based on farmers’ agricultural input orders in a given time period in order to make right inputs’ stocking decisions. The method in this paper successfully models and analyzes various machine learning classifiers that illustrate features which facilitate supplier’s decision before new inputs are stocked, which avoids losses hence leading to sustainable supplier business growth and development. We used classical metrics to compute the results of the different algorithms used and XGBoost, Random Forest, Decision tree, and K-NN all indicated a high percentage of overall dataset.