Prediction of Commodity Prices Using Supervised Machine Learning Models: The Case of MCX, India

The study investigates the application of supervised machine learning (SML) techniques to forecast commodity prices, utilizing eight years of data from the Multi Commodity Exchange (MCX). The study employs four SML models: Autoregressive Integrated Moving Average (ARIMA), Support Vector Regression (SVR), XG-Boost, and Long Short-Term Memory Recurrent Neural Networks (LSTM RNN). Further, the study focuses on predicting the prices of Mentha Oil, Crude Oil, Gold, and Copper for future periods. Among the models, LSTM RNN consistently delivered the highest accuracy in price predictions, outperforming ARIMA, SVR, and XG-Boost across all commodities. The superior performance of LSTM RNN is attributed to its ability to capture long-term dependencies and complex temporal patterns, making it particularly effective for the volatile nature of commodity markets. The findings demonstrate that LSTM RNN achieved the lowest error rates and the highest correlation with actual market prices, underscoring its robustness for practical applications in commodity price forecasting. ■ Keywords: Supervised machine learning, Multi commodity exchange, Commodity prices, Error rates and long-term dependencies