Enhancing Rainfall Forecasting Using a Hybrid Long Short Term- Particle Swarm Optimization (LSTM-PSO) Method

Heavy or protracted rainfall can lead to overflowing of rivers, streams, and low-lying areas. This can result in property destruction, infrastructure disruption, and displacement of populations. Conventional statistical approaches frequently rely heavily on past data patterns and may struggle to adapt to sudden changes or inconsistent weather conditions, leading to less accurate predictions during extreme events. When compared to traditional statistical methods, deep learning and machine learning techniques offer more robust and accurate models, greatly advancing the subject of rainfall prediction. The objective of this work is to improve rainfall prediction accuracy in the Cuddalore region by applying an integrated strategy that combines Particle Swarm Optimization (PSO) and Long Short-Term Memory (LSTM) networks. For the region's agricultural planning, water resource management, and disaster mitigation to be successful, accurate rainfall forecasting is essential. The temporal patterns present in rainfall data are best captured by LSTM networks, while PSO refines model parameters to increase prediction accuracy. The suggested LSTM-PSO model's performance will be evaluated using correlation coefficients, Mean Absolute Error (MAE), and Root Mean Square Error (RMSE). This work intends to provide strong and accurate rainfall forecasts by combining several approaches, assisting Cuddalore stakeholders in making well-informed decisions.