Tversky-RideNN based Feature Fusion and Optimized Deep RNN for Stock Market Prediction

Abstract

The stock market prediction has been considered a significant time-series forecasting research domain. The accurate prediction of stock prices helps the investors to gain more profits and to make better decisions while selling and buying the stock products. However., due to stock data's inherent., dynamic., and non-linear characteristics., the prediction process is more complex and remains a challenging task. Hence., this research proposes an effective strategy for accurate stock market prediction using the newly proposed Adaptive Competitive Feedback Particle Swarm optimization (CFPSO) method. The proposed approach includes four steps: extraction of technical indicators., feature fusion., data augmentation., and stock market forecasting. The stock market prediction mechanism is done using Deep Recurrent Neural Network (Deep RNN)., where the network is trained using Adaptive CFPSO. Moreover., the proposed approach has achieved a minimum MAE of 0.113., minimum MSE of 0.095., and minimum RMSE of 0.213.