HRSR-SVM: Hybrid Reptile Search Remora-based Support Vector Machine for forecasting stock price movement

Abstract

The prediction of stock movements remains an arduous process because of the dynamic and volatile nature of the stock market. In recent times, numerous stock prediction approaches are developed to forecast future stock movements. However, they have not achieved satisfactory results in predicting future stock prices. The objective of the paper is to enhance the performance of stock prediction by the Support Vector Machine-based Hybrid Reptile Search Remora (SVM-HRSR) approach. The data gathered from the stock market are utilized to examine the performance of the proposed prediction model which is collected from the National Association of Securities Dealers Automated Quotations (NASDAQ). The unstructured stock market data is structured to increase the prediction performance using the preprocessing steps namely data minimization and min–max normalization. The proposed SVM-HRSR approach accurately determines stock market movements based on the predicted patterns. The performance influencing factors of the Support Vector Machine (SVM) such as kernel parameter and penalty parameter is tuned mechanically using the Hybrid Reptile Search Remora (HRSR) algorithm to improve the prediction accuracy. Some of the indicators such as Mean Absolute Percentage Error (MAPE), Correlation Coefficient $(C_{c} )$, Root Mean Square Error (RMSE), Hit Rate (HR) and Symmetric Mean Absolute Percentage Error (SMAPE) estimated the performance of the SVM-HRSR method. The experimental results inherit that the proposed method achieved less error rates as compared to existing methods.