Comparative analysis of RNN and LSTM models for Agriculture Patterns Classification and Predictions

Abstract

This paper explores the comparative performance of Recurrent Neural Networks (RNNs) and Long Short-Term Memory (LSTM) networks in agricultural applications, particularly in crop classification, yield prediction, and optimal harvest timing. Both models were trained on a diverse dataset containing climatic variables (temperature, rainfall, humidity), soil data, and crop-specific features to predict crop yield and forecast harvest periods. The results revealed that the LSTM model significantly outperformed the RNN model in terms of accuracy, precision, recall, and F1-score, with the LSTM achieving an accuracy of 91% and a Mean Absolute Error (MAE) of 8% for yield prediction. Additionally, the LSTM model effectively incorporated climatic data, optimizing the prediction of harvest timing. This study highlights the potential of deep learning in precision agriculture, emphasizing the importance of integrating environmental factors into predictive models for better decision-making in crop management. The research underscores the need for further improvements in model accuracy through the incorporation of real-time data and hybrid modeling approaches. Future work also suggests the integration of deep learning with IoT-based systems for real-time monitoring and adaptive farming practices.