MACHINE LEARNING-BASED MODELING AND PREDICTION OF FOOD PRICE INFLATION IN ETHIOPIA

Abstract

Food inflation is an essential element of total economic inflation, indicating the pace at which food prices rise over a defined timeframe. This issue has garnered significant attention in its wider economic consequences and international comparison, specifically in Ethiopia. Ethiopia’s food price inflation, marked by volatility from interconnected climate, geopolitical and global market risks, has historically been forecasted using traditional models. While accurate predictions are critical for stabilizing the economy; these methods struggle to capture non-linear dynamics and external shocks. This study advances the field by leveraging a Random Forest model to enhance forecasting accuracy, thereby supporting policies to mitigate food price risks and promote economic stability. Thus, this study aims to model and forecast food price inflation in Ethiopia using a Machine learning method. The data were transformed, standardized and split into training and testing sets to enhance the forecast accuracy of both the machine learning and time series models. The selected models were evaluated based on performance evaluation criteria, including root mean square error, mean absolute error, and mean absolute present error tests. Auto-Regressive Random Forest model outperforming benchmark models with a 55% reduction in root mean square error (7.78 vs. 17.30) and 83.7% explanatory power (R²). Results reveal self-reinforcing inflationary cycles, with forecasts (2025–2029) indicating sustained volatility (20–30% range) due to primarily linked to monetary factor, international commodity markets and lagged value of food price inflation. The period from 2000 to 2007Q2 exhibited mild and gradual increases, followed by a sudden structural break in 2008Q4, marked by heightened volatility. A sustained upward trajectory emerged from 2015Q1 onward, reflecting persistent inflationary pressures. By combining autoregressive lags with Random Forest’s non-linear modeling, this study offers a scalable framework for food inflation forecasting in developing economies, providing actionable insights for policymakers to mitigate food insecurity and macroeconomic instability. The study highlights the dominance of inflation expectations and advocates for policy measures to stabilize agricultural inputs, and strengthen monetary frameworks.