Stock Price Prediction Based On The PCA-SVM-GARCH Model

The stock price is influenced by the economic cycle,the financial policy,the international environment and so on.The future stock price is unknown,as well as,it is difficult to predict the stock price.Investors will also be disturbed by information asymmetry,big client manipulation and herd mentality.Therefore,it is particularly important to explore a way to predict stock price trend.In order to predict the trend of stock price accurately,the PCA-SVM-GARCH model is first proposed in this paper.The model is optimized and improved on the basis of support vector machine model,so it has the characteristics of dealing with multidimensional nonlinear data.When we constructing SVM model,we can use the prediction effect as a criterion to screen out the optimal kernel function.Then the cross validation method is used to traverse the parameters combination,so that the optimal combination of parameters is determined.As a result we can predict the price.This model not only using the principal component analysis method to reduce the dimensionality of input variables and eliminate the multicollinearity among the variables,but also adding stock market volatility information into the model through the GARCH model.Therefore,PCA-SVM-GARCH model overall improves the performance of SVM model.In this paper,the research object from the stock that random selected(002230.SZ)extended to stock index(000300.SH).Prediction correct rate and prediction error is the standard for measuring the model is good or not.From the result,we find the prediction accuracy and prediction error of PCA-SVM-GARCH model are better than SVM model,SVM-GARCH model and PCA-SVM model.The stability analysis of the new model is taken in this paper.The influence of stock selection,holding time and data interval to this model is considered.The simulation results show that the model has a stable prediction effect,which shows that the model has universal applicability,market practicability and stability feasibility.This model can bring some guidance and reference value for many investors and market regulators.