USING MACHINE LEARNING TO PREDICT FOOD PRICES

Abstract

This project presents a machine learning-based framework for predicting global food prices by integrating historical pricing data with macroeconomic indicators. Addressing the challenge of food price volatility, the system uses public datasets from sources like the World Bank and Eurostat. Through careful preprocessing, feature engineering, and model testing, several predictive approaches were evaluated, including XGBoost, LightGBM, ARIMA, Prophet, and N-BEATS. Among these, Prophet demonstrated strong performance in capturing seasonal trends and producing accurate long-term forecasts. A hybrid Two-Stage Forecasting strategy was also employed to simulate future economic conditions and enhance prediction accuracy. The final outputs are visualized using Power BI dashboards, enabling intuitive exploration of trends across over 60 countries. This research highlights the practical potential of open data and machine learning in supporting policy decisions and mitigating risks associated with food price fluctuations.