Research On Quantitative Stock Selection Strategy Based On GRA-SVM

In terms of precise operation of trading timing and trading amount,the advantages of quantitative stock selection are very prominent.Overseas investment experience for decades shows that quantitative stock selection has stable performance and high returns,which is favored by investors.After entering the 21 st century,due to the rapid development of computer technology and the continuous improvement of quantitative theory,quantitative stock selection is becoming more and more popular.Quantitative stock selection strategies can be divided into basic quantification and market behavior quantification.Machine learning algorithms can play a role of classification or prediction in both basic quantification and market behavior quantification.Machine learning algorithm has outstanding advantages in dealing with non-linear and high-dimensional data feature problems,such as SVM.With the rapid development of computer science,machine learning algorithm is expected to play an increasingly important role in the field of quantitative stock selection strategy.With the continuous and rapid development of information technology,the price of financial products affected by complex factors can be quantified,and the accuracy is also rapidly improved.This has led to the upsurge of research and application of financial quantitative technology and automatic trading strategy.In order to improve the accuracy and practicability of financial quantification,this paper predicts the rise and fall of stock price based on GRA-SVM and designs an automatic trading strategy for quantitative stock selection.When forecasting stock price fluctuation based on GRA-SVM,the correlation degree between stock price and related factors is calculated based on GRA.Select the correlation coefficient p=0.5,calculate the correlation degree between the stock price and the related factors selected from Shanghai and Shenzhen 300(HS300)component stocks and rank them from large to small as sales cash ratio,market sales ratio,average net price ratio,net interest rate,transaction amount,total asset growth rate,asset return rate,price-earnings ratio,stock price skewness,total asset return rate,operating profit growth rate,net asset return rate,wave.Median amplitude,relative strength in December,10-day turnover rate,relative volatility of turnover rate,psychological line index,capitalization rate.Secondly,when classifying stock price fluctuations based on SVM,the factors which have little correlation with stock price are eliminated one by one,and the prediction accuracy and time-consuming under different combination of factors are obtained.As the number of selection factors decreases gradually,the time consumption of SVM prediction decreases,but the accuracy decreases.By balancing the prediction time-consuming and accuracy,the best factor is selected.The empirical results verify the validity of GRA-SVM selection factor and stock price prediction.The empirical results show that the accuracy of SVM in predicting stock price rise and fall is the highest when the five most relevant factors of stock price are the minimum sales cash ratio,market sales ratio,average net Market rate,net interest rate and transaction amount,which is 0.68.The vector space of factor data is non-linear.Therefore,the five factors are determined as the best combination of factors,which can be used to quantify the SVM in stock selection strategy to predict the rise and fall of stock prices.Finally,based on SVM,the stock price is forecasted according to the five best factors,and on this basis,the quantitative stock selection trading strategy is designed.Quantitative stock selection strategies include buying and selling criteria and frequency of strategy implementation.The basic strategy of stock selection in this paper is to buy some stocks which are predicted to rise in price and are not currently held,sell those stocks which have obtained high returns and buy the stocks which are predicted to rise in price again,and design risk control strategies.The results of stock selection strategy test without SVM kernel function are compared with those of stock selection strategy test with RBF kernel function of SVM.When the RBF kernel function is used in SVM,the results of quantitative stock selection strategy are the best.From January 1,2016 to May 1,2019,the annual return rate is 8.8%,which is significantly better than the basic annual return rate.The maximum withdrawal rate is 16.3%,and the volatility rate of return is 10.3%.At the same time,the results of uncontrolled risk strategy test and control risk strategy test using RBF kernel function of SVM are compared.Adding risk control strategy will reduce the annual return rate,but at the same time reduce the maximum withdrawal and return volatility,that is to say,reduce the stock-holding risk of stock selection strategy.The effectiveness of risk control strategy has been verified to some extent.