Predicting Food Prices Using Machine Learning and SHAP: A Focus on Four Essential Commodities in Bangladesh

Bangladesh is a developing country that relies heavily on agriculture. The availability and affordability of basic food items are very important for its people. Changes in food prices directly affect household expenses, food security and the overall economy. This study aims to predict the prices of four essential food commodities using economic and environmental data from 2004 to 2024. The commodities are rice, oil, wheat and lentils. Six machine learning models were developed and evaluated on MAE, RMSE and $\mathbf{R}^{\mathbf{2}}$. Gradient Boosting had the highest $\mathbf{R}^{\mathbf{2}}$ of 0.9697 for rice and 0.8755 for lentils, while Random Forest achieved 0.9337 for oil and CatBoost achieved 0.9073 for wheat. SHAP analysis was performed to explain the model's predictions. It showed that variables such as GDP, inflation, soil moisture, wind speed, and temperature have a significant impact on price changes. The results also showed that Bangladesh's economy and weather have a greater impact on locally produced foods than on imported goods.