Machine Learning-Based Forecasting of Headline Inflation in Ethiopia: an Empirical Comparative Study

Abstract Inflation is a critical indicator of national welfare and a major economic challenge globally, especially in Ethiopia. Traditional models have often been used to forecast headline inflation, yet they frequently struggle with accuracy due to the complex, nonlinear nature of the data. This study aimed to model and forecast headline inflation in Ethiopia using supervised machine learning approaches based on quarterly data from 2000 to 2023. This data obtained from the Central Statistics Service, National Bank of Ethiopia, Ethiopian National Meteorological Agency, and the World Bank. After analysis model performance was evaluated using RMSE, MAE, and MAPE. LASSO regression effectively addressed multicollinearity and achieved strong results (in-sample RMSE = 0.110; out-of-sample RMSE = 0.172). However, a nonlinear neural network model (NNAR) outperformed all other models, achieving perfect in-sample accuracy and the lowest out-of-sample error (RMSE = 0.128; MAE = 0.067), with 89.42% forecast accuracy. The predictors such as food and non-food inflation, import/export prices, political stability, exchange rate, transport data, rainfall, oil price, investment, unemployment, and agricultural prices were significantly affect headline inflation. The study demonstrated that machine learning, particularly NNAR, offers superior forecasting performance over traditional methods, providing better tools for economic planning and policy decision-making.