Construction and Exploration of a Strategy Portfolio Combining MACD and Monte Carlo Driven by Dual-LSTM

This study selects the CSSE 50 Index Dataset from 2020 to 2023 as the sample dataset, and picks 23 features to construct the predictors. During the data preprocessing, the Z-score standardization method is adopted, which improves the comparability of the data and the convergence speed of model training. After data cleaning, principal component analysis (PCA) was applied for screen out the key predictive factors. The PCA transformed the original variables into principal components, accounting for 92.3 % of the variance. To enhance the robustness of the model, the CRITIC method and random forest regression are combined in a weighted fusion manner to screen the key features, and this is verified through sensitivity analysis. When making predictions based on the stacked Long Short-Term Memory (LSTM) model, the study determines the signals in combination with the Moving Average Convergence Divergence (MACD) strategy, and uses the Monte Carlo method to simulate various combinations, improving the accuracy and reliability of the predictions.