Data mining approach for building predictive model for crop yield data

Abstract

The application of computer science has led to advancements in various sectors of economies including agricultural production, manufacturing and marketing. Computer algorithms have been used for prediction. There has been immense interest and research on crop yield prediction aimed at addressing food security. This has been achieved through the development of various crop models. Some researchers have studied yield prediction by applying computer science solutions. However, critical issues related to agricultural output have not been well addressed. This study looked at issues related to crop production. By using a case study approach and knowledge discovery data mining process this study was preceded by literature review, followed by analysis of daily 1950-2016 meteorological and annual 1950-2016 crop produce data in Njoro Sub- County (Rift Valley, Kenya). This study aimed to predict the agricultural outcome during per-harvest season by building predictive models of Artificial Neural Networks (ANNs), Decision Tree (J48) and PART rule via WEKA and evaluating the effectiveness and the accuracy of the built models. Results showed that ANN that was built through MLP function has the highest accuracy followed by Decision tree. The obtained results could help decision makers for achieving food security and the country’s productivity for the upcoming years continuously.