A Study of Cost-Sensitive SVM based on Region Labeling Method in Stock Prediction

Abstract:

In this paper, the region labeling method is proposed for the shortcomings of single point labeling method in traditional stock forecasting. The region labeling method can provide more useful information for training classifier and alleviate the problem of class imbalance to a certain extent, which is also more suitable for practical needs. At the same time, this paper constructs an RCS-Trader model, which uses cost-sensitive support vector machines and F_S measure to optimize. Compared with traditional stock predicting methods, RCS-Trader model works better and has higher return rate of investment.