Incorporating Transformers and Attention Networks for Stock Movement Prediction

Abstract

Predicting stock movements is a valuable research field that can help investors earn more profits. As with time-series data, the stock market is time-dependent and the value of historical information may decrease over time. Accurate prediction can be achieved by mining valuable information with words on social platforms and further integrating it with actual stock market conditions. However, many methods still cannot effectively dig deep into hidden information, integrate text and stock prices, and ignore the temporal dependence. Therefore, to solve the above problems, we propose a transformer-based attention network framework that uses historical text and stock prices to capture the temporal dependence of financial data. Among them, the transformer model and attention mechanism are used for feature extraction of financial data, which has fewer applications in the financial field, and effective analysis of key information to achieve an accurate prediction. A large number of experiments have proved the effectiveness of our proposed method. The actual simulation experiment verifies that our model has practical application value.

LEET: stock market forecast with long-term emotional change enhanced temporal model

“…They compared this approach with a conventional deep learning model and found that the Transformer model had superior predictive accuracy in all experiments and a better NAV profile. Li et al (2022) proposed a stock movement prediction model based on a Transformer and attention mechanism, and used social media text and stock price data including Twitter for experimental validation, by deeply extracting features of text and stock price, and using attention mechanism to obtain key information. The experimental results show that the method outperforms other baseline models in several metrics and has practical application value.…”

Integration of Emotional Factors With GAN Algorithm in Stock Price Prediction Method Research

“…The comparison models include: CNN, BiLSTM, CNN-BiLSTM, GRU, GAN, TEA and GAN-HPA models.In the GAN model, BiLSTM is selected for the generator, and CNN is selected for the discriminator.The experiments use the MAE, MSE, and RMSE as the evaluation indexes of the models, and the specific experimental values are shown in Table 7. TEA [14] 0.0231 0.0012 0.0359 GAN-HPA [20] 0.0224 0.0012 0.0348 TK-GAN 0.0200 0.0009 0.0306…”

“…Zhang [13] combined the empirical modal decomposition on top of the GRU model with attention mechanism to further improve the experimental results. Li et al [14] proposed a Transformer-based Attention Framework (TEA), which is a network framework for deep extraction of financial data features and further integration. It consists of a feature extractor and a cascade processor, and uses the research text and stock price information of five calendar days (calendar days in the original English) from the target trading day as the training data, which reveals the correlation between the research text and the financial data that are strongly correlated.…”