Research On Quantitative Stock Selection And Quantitative Timing Strategies Based On Deep Learnin

In the financial field,more and more investors pour into the market,and there are more and more researchers in the field of quantitative investment at the same time.Quantitative stock selection and quantitative timing are two important research contents in the field of quantitative investment.If a good strategy is formulated,the risk of investors’ investment can be reduced and a more stable income can be obtained.This paper selects CSI 300 as the stock pool,the factors and stock data from 2017 to 2020 as the training and verification set of the model,and the data from 2021 are used to test the model and strategy back test to study the quantitative stock selection and quantitative timing strategy.In the part of quantitative stock selection,first clean the data and reduce the dimension,and then build a multi factor stock selection model based on A-GRU.Select the stocks with the monthly return of each prediction result as the top 30 to build a portfolio,and compare the classification prediction effect with the GRU model.According to the result comparison,it is found that the classification result of A-GRU is more accurate.Due to the addition of attention mechanism,a series of weights can be simulated for the input data of the model,The model strengthens the extraction of important information and weakens the impact of non key information,so as to better identify the stock portfolio with higher return in the stock market,and the back test performance has higher return and lower risk.In the quantitative timing part,the daily rise and fall trend of constituent stocks is predicted according to the factor data.Compared with the classification prediction effect of GRU model classified according to the same method,it is found that the A-GRU model is more accurate in the classification of rise and fall.Finally,based on the above models and strategies,three strategies are established for comparison.The first stock selection strategy is to select stocks at the beginning of the month to build an investment portfolio,and then the annualized rate of return is 27.1%;The second timing strategy predicts the rise and fall of CSI 300 constituent shares in the next trading day,and then the annualized rate of return is 17.6%;The third is to combine stock selection and timing strategy,select stocks by month,and then conduct natural timing operation for the selected stock portfolio every month,and the annualized rate of return can reach 35.3%.The three strategies are back tested.According to the trend of the net worth curve of the strategy and the comparison of Sharpe ratio and Kama ratio,the combination of the final stock selection and timing strategy can obtain more returns than the single strategy,which is far higher than the benchmark return of CSI 300,and its ability to resist risks is strong,it can obtain higher returns with less risks,and the established portfolio can obtain excess returns,It has a good reference for investors to improve their income.