Predicting banana crop yield in Eastern Rwanda using machine learning models: insights from agricultural and meteorological data

This study predicts banana yields in Eastern Rwanda using machine learning. Data from the National Institute of Statistics of Rwanda and Rwanda Meteorology Agency were used, focusing on features like year, banana yield, organic fertilizer usage, rainfall, temperature, and relative humidity. Linear Regression, Random Forest, and Support Vector Machine (SVM) were applied to develop predictive models for banana crop yields. Random Forest delivered the strongest predictive performance (MAE = 294.00; RMSE = 372.86; R² = 0.93), outperforming Linear Regression and Support Vector Machine. The results confirm that the Random Forest algorithm is the most effective for predicting banana crop yields in Rwamagana District. This finding underscores the potential of deploying machine learning models for yield prediction, enabling informed decision-making and efficient resource allocation in banana farming practices across Rwanda. These insights are particularly beneficial for stakeholders such as local agricultural cooperatives, extension services, and government bodies like the Ministry of Agriculture and Animal Resources, aiding in the optimization of resources and the improvement of banana crop yields in Rwamagana District and similar regions in Rwanda.