CLSVP: A stock price prediction algorithm based on CNN-LSTM

Traditional time series models like RNNs face issues such as data loss and inability to maintain long-term memory when dealing with non-stationary and nonlinear stock data. The LSTM model, as an RNN variant, can handle nonlinear data and retain important time series information. In this paper, the CLSVP model, formed by combining the CNN and LSTM models, is used to model and analyze the constituent stocks of the CSI 400 Index in the Chinese stock market. Through empirical analysis of real stock return data, this study explores the learning and prediction performance of the CLSVP model in Chinese stock price changes and how its prediction results can be applied to quantitative trading strategies. It is found that the CLSVP model performs well in learning and predicting stock price changes, and the indicators based on its prediction results significantly improve the quantitative trading strategy.