Agriculture Performance and Inflation Dynamics in Rwanda: Application of Machine Learning

Abstract

This paper investigates the nexus between agricultural performance and inflation dynamics in Rwanda using machine learning techniques, particularly by applying elastic net regression to quarterly data spanning from 2008Q1 to 2023Q2. The study focuses on four key crops in Rwanda namely: maize, vegetables, Irish potatoes, and beans to assess the impact of dry spells and heavy rains during planting, growing, and harvesting periods on crop production. Results suggest that crops are more sensitive to heavy rains than dry spells. Additionally the study uses crop production, rainfall, and previous headline inflation (as a proxy for other costs) to predict fresh food inflation. As expected, the findings indicate that an increase in crop production lowers fresh food inflation, while the deviation from mean rainfall increases fresh food inflation. It was also found that high cost of production leads to a rise in fresh food inflation.Furthermore, the study uses six different Machine learning models and Auto regressive moving average model(ARMA) to forecast fresh food inflation ,it was found that Decision tree and Gradient boosting models outperform other models.Finally, the study uses the average of the two best models to forecast in-sample fresh food inflation, and the results are consistent with the Actual fresh food inflation.