Transparent Price Forecasting for Basic Food Commodities in a Developing Economy

Price fluctuations have been on a rise especially for the basic food commodities in developing economies like Uganda and are attributed to many causes. The use of machine learning algorithms was used to forecast the future prices in order to address the issue of price fluctuation. Several machine learning models were employed to implement the price forecasting model, including linear regression, KNN regression, support vector machine, XGBoost, light GBM, Random Forest regressor, and CatBoost regressor. The accuracy of each model was evaluated to determine the most precise and reliable one, which would be utilized for predictions. The models were further fine tuned to improve their accuracy and using hyper parameter tuning such as grid search. To enhance the explainability and interpretability of the models, Explainable Artificial Intelligence (XAI) techniques were employed. Specifically, LIME and SHAP were utilized to explain the decision-making process of the black box machine learning algorithms. Furthermore, the models were evaluated using evaluation metrics such as mean absolute error, mean squared error, root mean squared error and r-squared score upon which we selected the best price food forecasting model. According to the results, the XGboost outperformed all the other models in terms of performance and was selected to be deployed into the web application.