Crop Production Predictive Model Decision Support System

Agriculture is the backbone of the Ethiopian economy and contributes the highest percentage to the country's GDP. Among the agricultural sectors, crop production generates the highest income for most smallholder farmers across all regions of Ethiopia. The objective of this research is to build a model that predicts crop productivity and implement a decision support system. To achieve this, a hybrid Knowledge Discovery Process model was adopted. The datasets for this research were obtained from the Central Statistical Agency of Ethiopia, and a total of 25,000 instances were used for training and model development. For building the model and implementing the decision support system for predicting crop productivity, the WEKA data mining tool and Java NetBeans IDE were used, respectively. To meet the objectives of this research, various experiments were conducted using J48, HoeffdingTree decision trees, and PART rule-based classifiers. Additionally, the predictive performances of the classifiers were evaluated and compared using accuracy rate, confusion matrix, and ROC curve. Based on these evaluations, the PART rule-based classifier outperformed the others, achieving an accuracy rate of 95.44% and an ROC score of 0.992. Consequently, the PART rule-based classifier was selected to implement the model for predicting crop productivity. The experimental results of this research show that the main determinants of crop productivity include the main season (season type), use of extension programs, fertilizer usage, and fertilizer type. Therefore, the findings of this research are crucial for data-driven decision-making by policymakers and agricultural experts. They can use these insights to address factors that affect crop productivity and take corrective actions when necessary.