A Two-step Method Applying Support Vector Machine for Investment Decision

A two-step method is proposed in this paper to provide scientific basis for investment decision. The basis of this method is to construct feature pattern vectors as the input vectors of improved support vector machines (SVM) to map into class space. The first step is aimed at selecting stocks worth being invested, and the second step focuses on forecasting potential trend of the selected ones in the near future. Most notably, the evolving design of category labels ensures more detailed simulation of the trend. The two-step method is developed to produce appropriate models, improve the efficiency of decreasing capital risk, and meanwhile improving the investment yields for the business enterprises and individuals. Through simulations of Chinese A-share stock markets, experimental results obtained verify the two-step method is a promising and effective approach for investment decision-making, and the accuracy rate of forecasting reaches high levels.