Meta-Heuristic-Based Hybrid Resnet with Recurrent Neural Network for Enhanced Stock Market Prediction

Abstract

This paper is to design a new hybrid deep learning model for stock market prediction. Initially, the collected stock market data from the benchmark sources are pre-processed using empirical wavelet transform (EWT). This pre-processed data is subjected to the prediction model based on hybrid deep learning approach by adopting Resnet and recurrent neural network (RNN). Here, the fully connected layer of Resnet is replaced with the RNN. In both the Resnet and RNN structures, the parameter is optimized using the probabilistic spider monkey optimization (P-SMO) for attaining accurate prediction. When analyzing the proposed P-SMO-ResRNN, it secures 6.27%, 12.26%, 15.13%, 13.61%, and 14.10% more than RNN, DNN, NN, KNN, and SVM, respectively, regarding the MASE analysis. Hence, the proposed model shows enhanced performance. With the elaborated model and estimation of prediction term based on several analyses, this work supports the stock analysis research community.