Research on Quantitative Stock Selection Strategy Based on CNN-LSTM

Deep learning is widely used in the financial field of stock price trend predicting as an effective forecasting method. In this paper, we combine two neural network models of CNN and LSTM to realize the function of predicting the rise and fall of stock price, and control the risk at the same time. This paper preprocesses the original daily frequency trading data of all A-share from 2010 to 2021. We use the sliding window rolling method to generate data samples to train the neural network model, and predict the stock price of the stock data with the same characteristic factors through different network models. By changing the combination of model to set the contrast test, the results show that the rise and fall probability of the stock price based on the output of CNN network can be used as an effective method for quantitative stock selection, while the rise and fall probability of the stock price based on LSTM network can be used as an indicator of risk control, and the combination of the two can obtain excess returns, which has investment reference value for investors in stock selection.