An Innovative Approach to Short term Inflation Forecasting in Rwanda

Abstract

There is broad consensus that accurate forecasting of inflation is critical for effective economic planning and policy formulation. This study seeks to enhance the short-term forecasting capabilities of the National Bank of Rwanda (NBR) by applying a range of machine learning models to predict key components of the inflation index: Core, Food, and Energy inflation. The study assessed several machine learning algorithms, including Decision Tree, Random Forest, Gradient Boosting, K-Nearest Neighbors (KNN), Support Vector Regression (SVR), Elastic Net, and XGBoost. A comparative analysis revealed that Elastic Net regression consistently outperformed the other models in forecasting inflation components. Furthermore, when compared with the existing Near Term Forecasting (NTF) system used by NBR for short-term forecasts, Elastic Net regression showed superior performance. Based on these findings, the study recommends that the National Bank of Rwanda adopt a hybrid model to significantly enhance the accuracy of short-term inflation projections. The study recommends to the National Bank of Rwanda to explore these advanced modeling techniques to improve the Bank’s economic projections and decision-making processes.