Crop price forecasting using a Temporal Fusion Transformer for Krishna district of Andhra Pradesh

Indian farmers experience ongoing income volatility from fluctuating market prices, which erodes their financial health and long-term sustainability. To resolve this issue, a revamped Temporal Fusion Transformer (TFT), a state-of-the-art deep learning model is proposed for time-series forecasting with applications in agricultural price prediction. The TFT takes into account critical factors, including rainfall and temperature, previous price trends and market pressure in giving accurate, actionable forecasts to farmers. The model performed well when trained on a large database from Krishna district, Andhra Pradesh, India between January 2017 and September 2024. The TFT model achieved a RMSE of 99.13, MAPE of 2.16%, MAE of 72.08 and an accuracy rate of 93.24%. The system also allows the farmer to compare the forecast with the MSP and give them a very precise suggestion to maximise their revenue. It allows for farmers to take proactive decisions and supports an aware decision making approach by mitigating price volatility and stabilizing income.