Time Series Prediction using Laplace Transform-based Deep RNN- LSTM Approach

Abstract

This study presents a novel hybrid framework employing integrated Laplace transform-based Deep Recurrent Neural Network-Long shortTerm memory (DRNN-LSTM) model for time series prediction. In the presented work, the Ebola Optimization search (EOS) optimization algorithm was utilized to train the unsupervised DRNN-LSTM model. The Laplace transform in the proposed work eliminates the noises or unwanted features and helps to transform the input signal into a required format. Finally, the experimental analysis was performed and validated with the conventional techniques. The experimental analysis demonstrates that the designed model effectively eliminates the noises and also extracts the features accurately and also outperformed all the state-of-art techniques.