A New Stock Selection Model Based on Decision Tree C5.0 Algorithm

Abstract

Due to the disordered characteristic and strong randomness of China's stock market, the typical data mining algorithms currently used to analyze and forecast the stock have imprecise prediction outcomes. In order to solve this problem, based on the industry rotation cycle theory, this paper constructs a new stock selection model combining Decision Tree C5.0 Algorithm and factor analysis. Industry rotation cycle theory aims to analyze the development trend of various industries to find promising industries as initial stock pool. According to this principle, this paper selects four industries and the A-share stocks of these industries are used as initial stock pool. This paper builds a stock index system consisting of six effective factors based on the factor analysis of stocks financial indicators and technical indicators. Then Decision Tree C5.0 Algorithm is presented to realize the prediction of stock returns and the classification of stocks. The empirical test of the proposed stock selection model, using the data from the second and the third quarter of 2017 in China A-share stock market, demonstrates that this model has significant difference in the classification accuracy between low-yielding stocks and high-yielding stocks in that case classification accuracy shows a trend opposite against stock return rate. In a conclusion, this model can effectively help investors to avoid risks and make rational investment but has little effect on obtaining excess return.