PREDICTING RAINFALL INTENSITY USING SAE-LSTM AND SAE-BILSTM MODELS: A CASE OF KATHMANDU VALLEY

Abstract

Daily rainfall forecasts play a crucial role in enhancing agricultural productivity and safeguarding water resources, food supplies, and communities from the risks of floods and landslides. Accurate rainfall prediction is essential for mitigating the impacts of droughts and floods. Deep learning is increasingly recognized as a powerful tool for improving rainfall forecasting performance. This study aims to assess the effectiveness of hybrid deep learning models in predicting daily rainfall intensity. It approaches rainfall prediction as a 7-class classification problem, employing SAE-LSTM and SAE-BiLSTM models, which are then compared with standard LSTM and BiLSTM models. The models were evaluated based on metrics such as Accuracy, Precision, Recall, and F1-Score. The results indicated that the SAE-LSTM model outperformed the LSTM model, and the SAE-BiLSTM model outperformed the BiLSTM model. Among the hybrid models, the SAE-BiLSTM achieved superior performance, with an Accuracy of 84.60%, Micro Precision of 84.60%, Macro Precision of 84.85%, Micro Recall of 84.60%, Macro Recall of 84.99%, Micro F1-Score of 84.60%, and Macro F1-Score of 84.92%."