Incorporating investors’ stock review feature vectors in stock prediction models to improve the accuracy of stock price prediction. A stock price model STSB based on STL decomposition algorithm is proposed, which integrates the investor’s stock review feature vector constructed by the BERT model and the number of reads and comments of the stock reviews as features with the stock price features extracted by the STL decomposition algorithm through the attention mechanism to make stock price prediction. Comparing with the experimental results of five models, namely, KNN, ANN, DT, SVR, and RF, on six types of stock datasets, the results show that the accuracy of the STSB model in predicting the six types of stock prices is significantly improved compared with other single benchmark models. The STSB model with the addition of STL decomposition algorithm is analyzed through experimental validation to have better prediction effect and better prediction ability for the future stock price trend.
本站僅提供期刊文獻檢索。
