Stock Price Prediction Based on Data Mining Combination Model

To-Han Chang, Nientsu Wang, Wen-Bin Chuang

OAI: oai:igi-global.com:296707 • DOI: 10.4018/JGIM.296707

Predicting stock indexes is a common concern in the financial world. This work uses neural network, support vector machine (SVM), mixed data sampling (MIDAS), and other methods in data mining technology to predict the daily closing price of the next 20 days and the monthly average closing price of the future expected daily closing price on the basis of the market performance of stock prices. Additionally, by the mutual ratio of weighted mean square error the study achieves the best prediction result. Combining value investment effectively with nonlinear models, a complete stock forecasting model is established, and empirical research is conducted on it. Results indicate that SVM and MIDAS have good results for stock price forecasting. Among them, MIDAS has a better mid-term forecast, which is approximately 10% higher than the forecast accuracy of the SVM model; Meanwhile, SVM is more accurate in the short-term forecast.

Global Information Technology Computer Science and Information Technology IT Research and Theory