Quantitative Timing Strategy Using Support Vector Machine Based On Investor Sentiment

The computer information technology has advanced by leaps and bounds since the 1970s.With the support of statistical theory,the quantitative investment technology relying on computer technology has been developed rapidly.After 50 years of exploration,quantitative investment has accounted for more than 30%of the global financial investment goods so far,and it has become one of the main investment methods in the world's financial investment industry.For China which in the wave of financial globalization,the first public quantitative investment fund appeared in August 2004.Subsequently,private equity quantitative investment funds also developed rapidly,to the end of January 2021,the management scale of private quantitative investment fund has exceeded RMB 500 billion.At the same time,the rush to learn quantitative investment technology has begun to emerge from universities to major fund companies.There are multiple quantitative trading and learning platforms appeared on the Internet,such as JoinQuant?Datayes?RiceQuant?TradeBlazer.Quantitative trading has attracted more and more attention in Chinese financial market.According to the reality of Chinese stock market,from quantitative stock selection to quantitative timing,application of the machine learning technology that be widely studied,this paper construct a complete quantitative investment strategy,and has achieved a relatively good backtest rate of return.This paper selects the annual factor and daily agent sentiment data that the constituent stocks of CSI 300 Index from 2009 to 2020.First apply the multi-factor stock selection method,by using the multi-factor stock selection data from 2012 to 2019,by the principal component regression model to screen the top ten stocks according to the expected growth ability of the constituent stocks each year as The stock pool of next year to be selected,there are 10 stocks in total 80 stocks each year from 2013 to 2020 in the selected stock pool to be selected.Then,using daily proxy sentiment variables and principal component analysis to construct investor sentiment characteristics,after determining the label signals that need to be predicted,the investor sentiment characteristic data of the selected stocks each year are imputted into the support vector machine model(SVM).Predict the signal of each stock picked from 2013 to 2020,and trade based on the predicted signal each year.Finally,compare the cumulative return of each year's timing strategy with the cumulative return of the stock selection's buy-and-hold strategy and the cumulative return of the CSI 300 Index and backtested.The study showed that the support vector machine(SVM)that based on investor sentiment was used to quantitative timing strategy can obtain higher and more stable excess returns.