AgroAI: Revolutionizing Crop Marketing with Predictive Analysis

Abstract

Aims: To develop an intelligent, data-driven system AgroAI —that improves crop price prediction, market analysis, and decision-making for farmers and traders. The study aims to evaluate the effectiveness of the Temporal Fusion Transformer (TFT) model for multi-variate time-series forecasting and assess its impact on enhancing market transparency and crop marketing efficiency. Study Design: This study follows an applied experimental design involving machine learning model development, training, validation, and performance comparison against traditional forecasting approaches. Place and Duration of Study: The research was conducted within the Department of Artificial Intelligence and Data Science, over a period of eight months from February 2025 to September 2025, using publicly available agricultural datasets and state-level market records. Methodology: A dataset consisting of 10,200 market entries across four major crops (rice, wheat, soybean, and cotton) was collected. Features included historical prices, rainfall, temperature, soil moisture index, transportation cost, and local demand. Data preprocessing involved normalization, missing-value imputation, and correlation analysis. The Temporal Fusion Transformer (TFT) model was trained using an 80:20 train-test split, and its performance was compared with LSTM and XGBoost baselines. Evaluation metrics included Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and R² score. A market-insight module was also developed using feature-importance analysis for interpretation of factors influencing price fluctuations. Results: The TFT model achieved the highest predictive accuracy with an MAE of 4.7%, RMSE of 6.2%, and R² of 0.93, outperforming LSTM (R² = 0.86) and XGBoost (R² = 0.88). The analysis identified rainfall variability, transportation cost, and local demand index as the three most influential factors in price changes. The system generated market trend alerts with high accuracy, improving early-warning capability for farmers and traders. Conclusion: AgroAI demonstrates significant potential as a reliable, AI-driven tool for crop price forecasting and market analysis. By improving prediction accuracy and identifying key market drivers, the system can reduce financial uncertainty, support informed decision-making, and enhance the overall efficiency of agricultural marketing. Further integration with federated learning and real-time market feeds is recommended for broader deployment.