Crop Yield Prediction using Deep Learning Algorithm based on CNN-LSTM with Attention Layer and Skip Connection
Vijay H. Kalmani, Nagaraj V. Dharwadkar, Vijay Thapa
Abstract
Background: Accurate prediction of crop production is essential for efficient agricultural resource planning. Factors such as weather, soil moistureand temperature have a direct impact on crop yields, making precise forecasting vital. Methods: This study presents a hybrid model that enhances crop production prediction by integrating a 1D Convolutional Neural Network (CNN) with a Long Short-Term Memory (LSTM) network and an attention layer. The model is specifically applied to wheat and rice, major crops in India. The model evolves into a CNN-LSTM hybrid, designed to improve prediction accuracy by incorporating modifications, including multi-head attention and a multiplication skip connection. Result: When compared with conventional methods like Support Vector Regressor, Decision Tree Regressorand Random Forest Regressor, the proposed hybrid model shows significantly better performance. It achieves a Root Mean Square Error (RMSE) of 0.017, indicating low prediction error, a Mean Absolute Error (MAE) of 0.09 and a strong correlation between predicted and actual yields, with an R² of 0.967.