Employing Artificial Intelligence for Optimizing Crop Yield Forecasting in Bundelkhand Region

Abstract

Due to significant challenges in agriculture arising from climate variability, population growth, and resource limitations, precise crop yield prediction is essential to safeguard food security and promote sustainable agricultural practices. Therefore, accurate evaluation of crop production levels is crucial for effective agricultural resource management. The direct impact of variables such as temperature, soil moisture, and weather conditions on crop yields underscores the critical importance of precise forecasting in agricultural production. This paper presents a composite model that integrates a CNN (Convolutional Neural Network) with an LSTM (Long Short-Term Memory) network to improve the accuracy of agricultural output projections.    The model is specifically applied to rice and wheat, two of the most essential crops in India.  The proposed hybrid framework improves prediction accuracy using multi-head attention and multiplicative skip connections. Compared to more conventional methods, such as the Support Vector Regressor along with the Random Forest Regressor, the proposed hybrid model demonstrates significantly enhanced performance features. The Root Mean Square Error (RMSE) of 0.018 indicates negligible prediction error. Additionally, the Mean Absolute Error (MAE) is 0.08, and the R² value is 0.97, demonstrating a strong correlation between the predicted yields and the actual yields