Precise forecasting of crop prices is essential for improving food security and agricultural market efficiency. However, many existing models do not integrate advanced Deep learning methods with traditional statistical approaches. This study presents a hybrid model that enhances the accuracy of Wholesale Price Index (WPI) predictions for crops. The model combines ARIMA and LSTM (A-LSTM) with XGBoost for lag feature selection. The approach includes thorough data preprocessing, addressing seasonality and weather factors to boost forecast accuracy. XGBoost is used to select optimal lag features, while ARIMA and LSTM independently predict time series data. The hybrid model performs better by combining ARIMA’s statistical efficiency and LSTM’s deep learning strengths. Experimental results show that it reduces the Mean Squared Error (MSE) to 0.038, outperforming both ARIMA (MSE: 0.050) and LSTM (MSE: 0.086) models. To facilitate practical use a website has been developed. In addition to making the predictions easily accessible, this online platform gives farmers the ability to base their decisions on the data that has been anticipated. The model’s usefulness is further increased by combining cutting-edge methods with real-time data. The research shows how hybrid forecasting techniques can enhance agricultural price prediction systems, improving decision-making and offering stakeholders insightful information.