Ensemble Model for Stock Price Movement Trend Prediction on Different Investing Periods

Precisely predicting the stock price movement trend is essential for investors to gain enormous profit. However, due to the complicated financial environment, it is challenging to make an accurate market prediction. Machine learning is an effective tool to make such predictions. To build an effective investment strategy, selections of input features, prediction model with high accuracy, and buy-and-sell signals are very important factors. In this paper, we firstly screened out training features that are most relevant to stock price movement by calculating maximal information coefficient (MIC), then built an ensemble prediction model named SRAVoting basing on three outstanding classifiers on stock price movement trend prediction (support vector machine (SVM), random forest (RF), and AdaBoost (AB)), and proposed stock buy & sell strategies for day-span, week and month investing periods. Chinese stock price indexes and technical indicators were used to validate the proposed model and strategy. Result showed that SRAVoting model achieved higher prediction accuracy than SVM, while not necessarily higher annualized return rate than SVM based buy & sell strategies. Overall, the SRAVoting based strategies on different investing periods may benefit investors with positive returns.