Research on Stock Forecasting Based on EWP-BiLSTM-AM Model

Stock price prediction is not only a prominent research focus within the field of finance but also serves as a valuable reference for stock traders and investors. To enhance the model's accuracy in predicting stock price turning points, this paper introduces a stock prediction model, EWP-BiLSTM-AM, which integrates the Elliott Wave Principle (EWP), Bidirectional Long Short-Term Memory Networks (BiLSTM), and Attention Mechanism (AM). Leveraging the EWP and utilizing the Zigzag indicator, this model extracts Wave Feature Parameters (WFP) from the stock dataset. These WFP, in conjunction with historical trading data, serve as input for the BiLSTM-AM model. The BiLSTM is employed to extract temporal features from the data and generate predictions. Simultaneously, the AM selects the most pertinent features from the wealth of feature information, thus reinforcing the model's feature extraction capabilities. In this paper, comparative experiments were conducted on 12 stock datasets using various models. The results of the experiments demonstrate that the proposed model in this paper exhibits the best predictive performance on the datasets, with an average improvement of 20.1% in RMSE evaluation metrics compared to the LSTM model.