A Data-Driven Machine Learning Approach for Agricultural Commodity Price Forecasting in India

Abstract

The agricultural sector plays a crucial role in sustaining the global economy; however, it faces significant challenges due to volatile markets and fluctuating crop prices. Such uncertainties impede the ability of farmers, traders, and policymakers to make informed decisions, resulting in increased financial risks and inefficiencies. To address these issues, this study presents a data-driven price forecasting model that integrates demand–supply patterns, historical crop prices, current market data, and price fluctuation trends to generate accurate and reliable price predictions. The system provides stakeholders with actionable insights: farmers receive guidance on resource allocation, crop selection, and optimal timing for harvesting and sales; traders benefit from improved inventory management and investment strategies; and policymakers can leverage the forecasts to stabilize markets and safeguard vulnerable participants. Beyond decision-making, the model promotes efficient resource utilization and supports sustainable farming practices. Additionally, it provides a strong foundation for the integration of advanced technologies, such as artificial intelligence (AI) and machine learning (ML), in agriculture.Experimental evaluation on tomato commodity prices demonstrates that the model achieves a Mean Squared Error (MSE) of 25.87, Mean Absolute Error (MAE) of 4.78, Root Mean Squared Error (RMSE) of 5.09, an Akaike Information Criterion (AIC) of 16.58, and an R² score of 0.72, indicating that the model explains 72% of the variance in the dataset. These results reflect a balance between accuracy and simplicity, with comparatively low prediction errors for commodity price forecasting. Furthermore, comparative analysis with baseline forecasting techniques such as ARIMA and simple moving average revealed consistently lower error scores, establishing the model’s efficacy for crops with relatively stable price trends.