Leveraging Machine Learning for Accurate Groundnut Price Forecasting in Tamil Nadu: An XGboost Approach

Abstract

Abstract Accurate forecasting of agricultural commodity prices is vital for enabling effective decision-making within the agricultural ecosystem. This study addresses the implementation of modern machine learning algorithms, including XGBoost, Automated Machine Learning (AutoML) utilising PyCaret, and Auto-ARIMA, for forecasting groundnut prices in Tamil Nadu. The research integrates key temporal information, such as day of the week, month, and lagged data, to boost the prediction performance. The findings illustrate the improved performance of the AutoML technique, with the Light Gradient Boosting Machine (LightGBM) model attaining the lowest Root Mean Squared Error (RMSE) of 516.511 and Mean Absolute Percentage Error (MAPE) of 5.1%. The feature significance analysis indicates the substantial effect of year and lagged factors on the XGBoost model's predictions. The study provides vital information for stakeholders in the agriculture industry, including farmers, traders, and policymakers, by delivering precise price estimates to promote informed decision-making.