Multi-factor Quantitative Stock Selection Strategy Design Based On Individual Investor Sentiment

As an important method of quantitative investment strategy,the multi-factor stock selection model has many problems in the timeliness of stock selection factors and the ability to deal with risks.On the one hand,the effectiveness of the factors needs to be further verified in the new market environment.On the other hand,the multi-factor stock selection model emphasizes the improvement of stock selection factors but lacks consideration of the risk of the large market,resulting in most models with strong stock picking ability,but weaker timing ability.Therefore,this thesis optimizes the traditional multi-factor stock selection model under the premise of considering the timeliness of factors.This thesis uses text mining to capture real-time postings of CSI 300 index stocks,constructs a financial-specific dictionary and uses text semantic analysis to construct daily,weekly,and monthly sentiment indicators for individual investors,and studies its prediction of stock market returns.The study found that the daily individual investor sentiment can positively predict the current stock market situation,the weekly individual investor sentiment can positively predict the current stock market situation and reversely predict the next stock market situation,the monthly individual investor sentiment can negatively predict the next stock market situation.Set the month in which individual investor sentiment is higher than the sum of the mean and the two standard deviations or lower than the difference between the two values as a possible high volatility range,and get the month in January 2016,February 2018,and the year 2018.Avoidance will be conducted in June,May 2019 and July 2021,that is,the position closing operation will be carried out one position adjustment period in advance.Taking each issue of CSI 300 constituent stocks as the sample stock pool,selecting the monthly data of the sample stock factors from January 2010 to November 2021,and using Python to perform single-factor detection,including: introducing 102 candidate factors to build a factor library,data preprocessing is performed on the sample data,that is,outliers,standardization,and market value industry neutralization.Factors are tested by IC method,regression method test,and group test method,after removing redundant factors,a multi-factor stock selection model is established by using the selected factors.Using the CSI 300 index as the benchmark portfolio,writing a backtest module through Python,and comparing and analyzing various indicators of the strategy and the benchmark portfolio,it is found that the constructed multifactor stock selection model can obtain higher excess returns than the benchmark portfolio,but it performed poorly on risk metrics.In order to further optimize the risk index of the strategy,the individual investor sentiment index was introduced into the multi-factor stock selection model as a timing strategy.The backtest shows that the introduction of individual investor sentiment index as a timing strategy can significantly optimize various indicators of the multi-factor stock selection strategy,of which the annualized excess return reaches 16.6%.Finally,the out-of-sample,position change cycle and position opening and adjustment delay tests are carried out on the strategy.The results show that the strategy has better performance in bear markets.The delay test should pay attention to the effects of the beginning of the month and the end of the year,and the strategy parameters should be set according to investors Set the preference for backtesting indicators and the grasp of market conditions.