AI BASED CROP MARKET PRICE PREDICTION SYSTEM USING MACHINE LEARNING

Abstract: This paper presents a machine learning based system for predicting the market price of agricultural crops to support farmers, traders, and agricultural policymakers in making timely and informed decisions. Agricultural commodity prices in India are highly volatile and are influenced by seasonal cycles, weather conditions, demand and supply fluctuations, transportation costs, and government policies. This volatility, combined with a lack of reliable price information at the farm level, frequently forces farmers into distress selling at unfavourable rates. The proposed system, CropCast AI, integrates historical market (mandi) price data, weather parameters, seasonal indicators, and supply demand factors to forecast future crop prices for a selected crop, market, and time period. Multiple regression and time series models including Linear Regression, Random Forest, XGBoost, and Long Short Term Memory (LSTM) networks were trained and compared on historical price datasets. The system is deployed as an interactive web application built using the Flask framework, allowing users to select a crop, state, market, and target date and instantly view the predicted price along with historical price trends. Experimental results demonstrate that the ensemble and deep learning models outperform classical baselines, with the best model achieving a coefficient of determination (R2) of 0.93, a Mean Absolute Percentage Error (MAPE) of 7.8%, and a Root Mean Square Error (RMSE) within acceptable agricultural forecasting limits. The proposed framework reduces information asymmetry, helps farmers decide when and where to sell their produce, supports better procurement and storage planning, and provides a low cost decision support tool for the Indian agricultural ecosystem.

Keywords: Crop price prediction, machine learning, agriculture, XGBoost, LSTM, time series forecasting, mandi prices, regression analysis, decision support system, precision agriculture.