TCN Stock Price Prediction Model Based on Channel Attention Mechanism

Since the stock price fluctuations of the concept stocks of the metaverse culture and media sector has the same sector correlation, this paper proposes an improved TCN (Temporal Convolutional Network) models based on the channel attention mechanism, which makes the model focus on effective feature channels. First draw the heat map of the Pearson correlation coefficient of the stock closing price, insert the channel attention module on the TCN basic model; then use the standardized method to process the sample data, unify the sample data dimension, and construct the data set. The experimental results show that the performance of the improved TCN model based on the channel attention mechanism on A shares and US stocks is the same as the RMSE (Root square error) of the prediction results of multiple stocks. Better than the TCN basic model, using the channel attention weights to correct each channel can help the model learn and converge better, and it is effective for the optimization of the prediction model.