Research on Stock Index Futures Price Prediction Using ATTVAE Sparse Transformer Model

The non stationarity, nonlinearity, and long-term dependence of the stock index futures market increase the difficulty of market prediction, making accurate prediction of market trends extremely challenging. Propose an ATTVAE Sparse Transformer model that combines an improved Variational Autoencoder (VAE) with a Transformer to achieve hybrid deep learning prediction of stock index futures prices through data augmentation strategies. Firstly, by using an improved variational autoencoder to extract features of the trend and periodicity of the raw data of stock index futures, a certain level of interpretability at the time level is provided for the data generation process; Secondly, by introducing a sparse multi head attention mechanism to reduce computational complexity and cost, an improved Transformer network is used to predict stock index futures prices. The results show that after adding the VAE enhanced stock index futures price data, the sparse long attention Transformer can learn more fully from the original stock index futures data, capture key feature information more effectively, and achieve more accurate predictions. The ATTVAE Sparse Transformer model was evaluated using root mean square error (RMSE) and mean absolute error (MAE) on data from the SSE 50 stock index futures, CSI 300 stock index futures, and CSI 500 stock index futures. It was found that the model outperformed other models in terms of prediction accuracy.