Portfolio Research Based on LSTM Model

In this paper, the LSTM model is used to predict the stock prices of the top 10 constituents of CSI 300 index and construct a portfolio. The empirical results show that the LSTM model performs well in predicting the CSI 300 index, and the difference between the predicted and actual values is relatively small. The robustness of the model is assessed by the time series cross-validation method, and the results show that the average prediction error of the model on different folds is very low, which indicates that the model performs well in dealing with the time series data and has high prediction accuracy and robustness. The analysis of the portfolio results shows that the portfolio is profitable on average over the time considered and can achieve higher returns than the CSI 300 index. This study provides a reference for investors to construct portfolios with higher returns and provides new ideas for the development of financial technology.