An Efficient Rainfall Prediction Using Improved Multilayer Perceptron

Data mining relies on the contemporary digital landscape to uncover previously unnoticed relationships within vast datasets. Time series analysis is employed to scrutinize patterns within data over defined intervals, particularly in predicting future events. This scientific approach is notably applied in forecasting activities over time, and rainfall prediction is a pertinent example. An advanced multilayer perceptron neural network has been introduced, utilizing intelligent techniques for time series rainfall prediction. The accumulated rainfall data was processed through this proposed network. Evaluation metrics, including mean squared error, maximum error, and normalized mean squared error, were employed to assess performance. Results indicate that, especially within a short timeframe, the proposed multilayer perceptron network outperforms other models. This assessment sheds light on the efficacy of these models, demonstrating that the projected neural network closely aligns with actual outcomes for desired yields in multistep ahead forecasts.