Stock Trend Prediction Based on SVM and PSO Parameter Optimization

摘要

In recent years, the use of support vector machine classification on the stock market forecasts is of great popular. However, forecasts using SVM to do classification needs to adjust relevant parameters (mainly the penalty parameter c and the kernel function parameter g) to achieve the ideal prediction accuracy of classification. Many researchers have tried to use cross-validation (CV, Cross Validation) ideas, to find the penalty parameter c and the kernel function parameter g. Using cross-validation proved although the idea can be obtained in some sense optimal parameters, you can avoid the occurrence of the state of over-fitting and lack-of-fitting, but the use of grid search of this non-heuristic optimization, and sometimes if you want to in a larger scale search for the best c and g within the parameters will be very time-consuming.