Application Of SVM Based On Investor Sentiment In Quantitative Investment

Since the formation of China's capital market in the 1990s,its scale and volume have presented a trend of rapid growth.However,accompanied by rapid growth,there have been many large fluctuations.In contrast,the mature capital markets of western countries did not show such violent fluctuations.The reason is that the immaturity of institutions and mechanisms is one of the important factors causing such differences.However,many market anomalies cannot be explained by traditional financial theories.Combined with the research of current scholars on market anomalies,the author combines traditional finance and behavioral finance,trying to reflect the operation of China's stock market more realistically from the perspective of investor psychology.Behavioral finance believes that investor sentiment is an important factor influencing the pricing of risk assets,and the irrational factor of investors is one of the bases for the difference between embedded value and market value.Therefore,from the perspective of investor sentiment,the study of the relationship between irrational fluctuations of investor sentiment and market changes will help us clarify market logic,effectively avoid systemic risks,enhance the effectiveness of supervision and thicken investor returns.The overall structure of this paper can be divided into the following four parts:The first part is the introduction,which mainly includes the research background,research significance and the structure of this paper,and points out the possible innovation points of this paper.The second part is a literature review.This part mainly reviews the definition and measurement of investor sentiment by domestic and foreign scholars,studies on the relationship between investor sentiment and stock market return rate,and introduces the research methods and theoretical knowledge of support vector machine used in this paper.The third part is the emphasis of this paper.Based on the existing studies on investor sentiment by domestic and foreign scholars and the differences between China's stock market and western mature capital market,this paper selects the proxy variables suitable for investor sentiment in China's stock market,and combines multiple proxy indicators into a composite index of investor sentiment using the PCA method.The wavelet denoising method was used to remove the noise of the investor sentiment index,and the index vectors that could better reflect the trend of investor sentiment were obtained.Finally,the VAR model and SVM model were used to respectively back-test the monthly return data of HS300 from May 2010 to December 2015.Through empirical analysis,it is found that investor sentiment is an important factor affecting index return rate,and the SVM model considering investor sentiment has obvious advantages over other models in terms of positive win rate,with higher cumulative gain and total win rate.By contrast,the VAR model,which takes into account the stationarity of time series,has a relatively high accuracy in the accuracy of negative prediction,so it also has a high total win rate and accumulated earnings.Finally,it is concluded that in the prediction of financial time series,the behavior and emotion of investors should be a significant influencing factor for us to consider.The introduction of machine learning into non-linear modeling will be a great analysis tool for investors engaged in quantitative investment.The fourth part is the summaryand prospect,and points out the deficiencies of this paper and the direction of future in-depth research.