Stock Trend Prediction Based on Improved SVR

Abstract

Forecasting the change trend of stocks and making investment decisions based on the prediction results can effectively avoid risks in stock investment and increase investment income. The combination of machine learning and big data provides an effective way for stock trend prediction. At present, many machine learning algorithms have been used for stock trend prediction and have achieved good results. In view of the fact that the support vector regression (SVR) algorithm has a large deviation from the prediction of individual stocks, this paper proposes an improved SVR algorithm. Different from traditional SVR, the enhanced SVR uses the recent data to train the model, rather than all history record. An experiment was performed to compared the proposed SVR with traditional SVR and decision tree algorithms. The RMSE are used as metric. The result shows that the proposed SVR is better than the existing comparing algorithm.