Research On Quantitative Investment Strategy Based On Emotion Analysis And Machine Learning

With the development of network technology,more and more individual investors express their views on the trend of the stock market or a certain stock through online platforms such as stock bars,and communicate with other investors.According to the theory of behavioral finance,investors’ irrational decisions may have a significant impact on stock returns.his paper constructs a new network emotion factor using the subjective emotions of investors on the network platform,and applies it to a multifactor quantitative stock selection model constructed using support vector machine(SVM)algorithm,improving the effectiveness of the multifactor stock selection model.Firstly,a crawler program is used to collect the stock review texts of 300 Shanghai and Shenzhen constituent stocks from Dongfang Fortune Stock Bar from July 3,2021 to July 3,2022.In order to improve the accuracy and efficiency of sentiment classification of stock review texts,the collected stock review texts are divided into long and short categories.Short texts are classified using naive Bayesian(NBM)models for emotion,while long texts are classified using short-term memory(LSTM)models for emotion.After classification,calculate the value of online emotional factors,and use a linear regression model to demonstrate that online emotional factors significantly affect stock returns in the next period.Then,financial factors and market factors are screened,and four factors are selected from them.Support Vector Machine(SVM)is input to train and backtest the quantitative stock selection model.Four factors and a total of five previously constructed network emotion factors were input into support vector machine(SVM)for training and backtesting of the quantitative stock selection model.Comparing the backtesting results,it was found that:(1)Compared with the quantitative stock selection model without network emotion factors,the earnings of the quantitative stock selection model with network emotion factors increased by11.23%.(2)The quantitative stock selection model with online emotional factors has lower beta coefficient,maximum pullback,and strategic volatility than the quantitative stock selection model without online emotional factors,while the Sharp ratio,information ratio,Sotino ratio,and Karma ratio evaluation indicators perform better.This indicates that the strategy achieves higher returns with lower risk.China’s stock market is not a fully efficient market,and investor textual sentiment can significantly affect stock returns.The online sentiment factor is an effective stock selection factor that can select stocks with high returns.Due to the addition of behavioral finance factors that reflect the subjective emotions of investors,namely,the online sentiment factor,to the stock selection model,the effectiveness of the quantitative stock selection model has been improved.