Rainfall prediction using time series data based on RSJS']JSO_BiLSTM

Time series data is highly dimensional and is updated continuously. Forecasting rainfall through time series analysis is a critical aspect of this application. The accurate prediction of rainfall is achieved using the Time series data, which are essential for the early warning of potential droughts and flooding events. In this research, an efficient rainfall prediction model is developed using time series data with a Bidirectional LSTM (BiLSTM), optimized through the Rat Swarm Jellyfish Search Optimizer (RSJSO). This model offers reliable predictions, providing timely and precise information to minimize the impact of extreme weather on communities and infrastructure. Here, the missing data imputation is employed for preprocessing the acquired data. Moreover, the technical indicator extraction is done to reveal the average prediction of rainfall over time and provide a clearer picture of support and resistance. Also, the feature fusion is effectively done using the chord distance and Rider Neural Network (RideNN), which offers better performance and low computational complexity. Moreover, data augmentation is done for fused features using an oversampling technique resulting in rainfall prediction. Here, the integration of Rat Swarm Optimizer (RSO) and Jellyfish Search Optimizer (JSO) forms the devised RSJSO. Also, the Rain in Australia dataset is used for analyzing the performance of the RSJSO_BiLSTM model. Furthermore, the performance of 0.125, 0.354, 0.122, and 0.015 are obtained by the RSJSO_BiLSTM by considering the evaluation indicators namely Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Percentage Error (MAPE), and Relative Absolute Error (RAE).