Enhancing Food Security through Seasonal Crop Yield in Nigeria Using Machine Learning Techniques

Formerly, farmers were faced with the challenge of knowing the yield from their farm. Prediction is very important in agriculture. In the past, yield prediction was performed by considering farmers’ experience with crops. The number of yields cannot be obtained accurately by taking into consideration yields from previous farming seasons. The traditional methods used by the farmers are very slow and unreliable, and large amounts of crops are damaged due to bacterial attacks, erosion and other natural factors. Adoption of big data and machine learning is a key tool to digitalise the agriculture sector and other sectors. Though there is a long debate on its applicability to agriculture, this study addresses how machine learning contributes to digital agriculture in terms of the prediction of crop yields. Three (3) different techniques have been adopted using support vector machine, random forest and decision tree classifiers for predicting crop yields during rainy and dry seasons after the datasets have been subjected to a series of cleanings to get the best in terms of yields and efficiency. The methodology adopted was Cross-Industry Standard Processing for Data Mining (CRISP-DM) with a series of stages before the final deployment. Southeast farming during the rainy season gives the best in terms of yields, accuracy and efficiency using a decision tree classifier with 98.61% accuracy.