Research And Implemention Of Intelligent Investment Strategy In Chinese Stock Market

Investment strategy research has always been a popular research direction in the field of financial engineering and computing technology.Specially,multi-factor stock selection strategy is the top priority in investment strategy.Since the three-factor model was proposed in 1993,the theory and technology of multi-factor stock selection strategies have been greatly developed.In recent years,with the rise of deep learning,the investment strategy of applying deep learning methods to quantitative stock selection has attracted much attention and achieved good investment results.Quantitative investment strategy in overseas markets is relatively mature,but it is still in their infancy in China and has greater research value.On the other hand,the amount of stock market data is huge,and manual processing is time-consuming,labor-intensive and difficult.Therefore,for the Chinese stock market,this paper uses the deep learning model to comprehensively mine market information,fit the relationship between stock returns and influencing factors.And the we build intelligent investment strategies based on the results of stock return scores,so as to more scientifically invest in stocks.At present,no researchers have applied TCAN to stock return prediction and stock selection.This paper designs Adaptive-Factors Temporal Convolutional Attention-based Network(AF-TCAN),which is based on Temporal Convolutional Attention-based Network(TCAN)and introduces Factors Attention(FA).FA weights the importance of input factors to remove factor noise,and then TCAN performs stock factor feature extraction.During the period from June 2018 to May 2019,the AFTCAN model proposed in this paper can effectively improve the classification accuracy of stock returns,reaching 59.95%.Then,using the prediction results of AF-TCAN to conduct stock selection backtest experiments in the domestic market,the results show that its cumulative yield curve is significantly better than that of CSI 300.At the same time,it is compared with the traditional linear model,LSTM model and Twobranch model.The cumulative rate of return and main risk indicators of AF-TCAN are higher than those of the above models.Its cumulative return rate is 22.48%higher than that of the linear model,23.59%higher than that of the LSTM model,and 26.15%higher than that of the Two-branch model;its information ratio is 58%higher than that of the linear model,63%higher than LSTM model,and 42%higher than Two-branch model.In addition,considering the impact of news events on the stock market,this paper proposes AF-TCAN with news model based on AF-TCAN,adding the news information channel.The experimental results show that its Sharpe ratio is 11%higher than that of AF-TCAN model without news information,which indicates the return on risk taking is relatively higher.Finally,this paper designs and implements an intelligent quantitative investment system to provide users with an intelligent multi-factor stock selection platform in the Chinese stock market.