Stock Prediction Model Based on Wavelet Packet Transform and Improved Neural Network

Abstract

With the advent of the era of big data, the network’s intervention in the stock market has deepened and the security of the stock market has been seriously threatened. In order to maintain the security of the stock market, this paper proposes a long short term memory prediction model based on wavelet packet decomposition and attention mechanism (Wav-att-LSTM). First, Wav-att-LSTM uses the XGBoost algorithm to select important feature variables from the stock data, and then uses wavelet packet decomposition to extract stock frequency features, which are used as the next input. Finally, the LSTM with the attention mechanism is used as the prediction model to predict the frequency component. This paper uses the stock dataset of the \\({S\\delta P500}\\) for performance verification. The experimental results show that Wav-att-LSTM has higher prediction accuracy and less hysteresis than some advanced methods.