Research On Stock Price Trend Prediction Based On Optimized Principal Component Support Vector Machine

With the steady development of China’s economy,more and more investors enter the stock market,stock price trend prediction has always been the focus of stock investors.At the same time,the status of data research continues to rise in the business and financial field.Many scholars have applied different theories and models to the stock market and achieved good results.In this paper,based on the relevant data of Zhifei Biological in 2020,this paper selects 17 indicators as predictive variables,including the before closing price,opening price,maximum price,trading volume,daily amplitude and daily turnover rate of tradable shares.Firstly,the k-nearest neighbor model,decision tree model,random forest model and support vector machine model are established according to the prediction variables.Among the above models,support vector machine model has the highest prediction accuracy on the test dataset,so tabu search artificial bee colony algorithm is used to optimize the parameters of this model.The prediction accuracy of the tabu search artificial bee colony algorithm optimized predictive variable support vector machine model on the test dataset is improved,which is 6.98 percentage points higher than the predictive variable support vector machine model.However,in general,the prediction accuracy of the above models is low and the number of predicted variables is large.Therefore,it is considered to extract the principal component of the predicted variables and then establish k-nearest neighbor model,decision tree model,random forest model and support vector machine model.It is found that the principal component support vector machine model has the highest prediction accuracy by comparing the above principal component models in the test dataset.Then build the principal component support vector machine model optimized by tabu search artificial bee colony algorithm,the model of the test dataset prediction accuracy is up to 72.09%.It is 13.95 percentage points higher than the predictive variable support vector machine model,6.97 percentage points higher than the predictive variable support vector machine model optimized by tabu search artificial bee colony algorithm,and 6.97 percentage points higher than the principal component support vector machine model.The principal component support vector machine model optimized by tabu search artificial bee colony algorithm performs better than other models in this problem.In this paper,a principal component support vector machine model for tabu search artificial bee colony algorithm optimization is constructed for the first time,tabu search artificial bee colony algorithm to optimize the two undetermined parameters of support vector machine model,solved the existing in the traditional support vector machine model of unable to determine the optimal parameters of the problem.In addition,tabu search artificial bee colony algorithm is an improvement of artificial bee colony algorithm,which not only has fast convergence speed,but also can avoid falling into the local optimal solution problem.Will finally the principal component support vector machine model optimized by tabu search artificial bee colony algorithm is applied to the dispatch of Jidong Cement,IFLYTEK,CITIC Securities,Kweichow Moutai four stocks.Compared with other models,the prediction accuracy on the test datatest of the principal component support vector machine model optimized by tabu search artificial bee colony algorithm has been improved to different degrees.According to the empirical results,the principal component support vector machine model optimized by tabu search artificial bee colony algorithm has excellent performance in stock price trend prediction,and its prediction accuracy is significantly higher than that of K-nearest neighbor model,decision tree model,random forest model and traditional support vector machine model on test dataset.This model can be used as a reference to study stock price trend based on statistical learning theory,and can also be a scientific and effective method for investors to predict stock price trend.