Residual temporal convolution network with novel activation function for financial prediction with feature selection procedures

Abstract

Finance provides a major contribution to countries economic growth. A deep understanding of the financial market helps to offer better financial returns in the future. The financial market generates more complications for predicting the complicated system dynamics. Different machine learning techniques are implemented to execute the financial market prediction and they didn’t provide better outcomes in financial returns. Predicting the stocks in the yearly phases brings huge profits to the stock market traders and helps to make better financial decisions. Different deep learning and machine learning techniques help to predict the accuracy of the stock market. The deep learning techniques effectively handled the enormous unsupervised and unstructured data. In order to achieve better results, the intelligent deep learning model is proposed for forecasting the financial crisis. Initially, the raw data are fetched from the benchmark datasets. Subsequently, the multi-objective-based feature selection process takes place, where the features are optimally selected by using the Updated Random Variable-based Coati Optimization Algorithm (URV-COA). Due to this selection, various constraints like correlation, relief score, and data variance are considered for formulation. Finally, the resultant features are subjected to the Residual Temporal Convolutional Network with a Novel Activation function (RTCN-NAF) for predicting the financial cost. Therefore, the experimentation for the proposed model is assessed by using the divergent metrics and compared with other traditional methodologies. On the contrary, the suggested work achieves better results that can prove the effectiveness of the prediction system.