Stock Price Model And Empirical Research Based On Independent Component Analysis

In the fast-changing stock market,we can judge the stock price reversal point earlier,so as to sell high and attract low,we can get rich investment returns when we can master the main factors affecting stock price fluctuation and the influencing ways.Therefore,the prediction of stock price reversal points plays an important role in the investment decision-making.The prediction of stock price reversal points and the analysis of influencing factors of stock price fluctuation based on independent component analysis are studied in this paper.The main contents are as follows:(1)Aiming at the problems of inaccurate calculation and easily falling into local minimum of some traditional stock price inversion point prediction methods,this paper proposes a stock price inversion point prediction model(ICA-SVM)which combines ICA and SVM.The model uses ICA to extract the structural characteristics of stock price fluctuation and rebuild the stock price,which highlights the characteristics of stock price reversal and weakens the influence of other interference factors.Meanwhile,SVM is used to realize the prediction of stock price reversal point.Empirical results show that ICA-SVM model achieves better results for bank stocks with smaller stock price changes,GEM stocks with larger vibration and real estate stocks with larger market share.Its recall rate,precision rate and F_Measure are improved compared with SVM model.(2)DIC A can overcome the disadvantage of ICA classification performance,but its computa-tional complexity is high.GWO is widely used in engineering optimization problems because of its simple structure,few parameters and fast convergence speed,and has achieved good results.Therefore,this paper proposes a reversal point prediction model based on GWO-DICA-SVM.The model uses GWO instead of the gradient descent algorithm in DICA.The calculation is simple and to some extent alleviates the problem that traditional methods are easy to fall into local minimum.Empirical results show that GW O-DICA-SVM model has achieved good results in predicting the reversal points of bank shares,GEM shares and real estate shares,and its F_Measure is nearly 10%higher than that of SVM model.(3)Noise often exists in stock price data.Noise removal is a key link in stock price analysis.ICA is used to detect and remove noise.Considering the strong collinearity between various factors,an analysis model of influencing factors of stock price based on ICA and LASSO regression is proposed.From the empirical results,we can see that the regression equation obtained by ICA-LASSO factor analysis model has reasonable explanations for the real estate,new energy and GEM stocks,which shows the validity of the method.