A Stock Prediction Model Based on CNN-BiLSTM and Multiple Attention Mechanisms

Abstract

Based on the traditional stock prediction research, this paper introduces bidirectional long and short-term memory and CNN models. The model extracts local and global features from stock data, extracts the most critical and complex features, and combines the attention mechanism to assign feature information weights to improve prediction accuracy. The empirical results show that the CNN-BiLSTM hybrid model combining the SE attention mechanism has the optimal prediction effect, and the R ² value reaches 0.9725, which can be used for practical applications.