Comparison of Stock Price Predictions Using Support Vector Regression and Recurrent Neural Network Methods

Abstract

Apart from fundamental analysis, there is also technical analysis in predicting stock price movements. From predictions based on conventional statistics, to machine learning and even deep learning, stock market players, be they stock owners or investment managers or stock consultants, use this to make the analysis more accurate and to profit from the analysis. M Asset as a company also wants its consultants to master machine learning and deep learning programs to be able to assist their customers in analyzing stock price movements. However, there are so many machine learning and deep learning methods and how to find out the performance of these two methods. For this reason, this study intends to compare the machine learning method represented by Support Vector Regression (SVR) with the deep learning method represented by the Recurrent Neural Network (RNN) with time series data from Astra Agro Lestari shares. With error comparison parameters and root mean square error (RMSE). After conducting research, it turned out that the RMSE of the SVM was 132.42 and the RMSE RNN was 354.86. which means that the error rate of RNN is higher than SVM and the accuracy level of SVM is higher than RNN