A Review of the Application of Transformer in Financial Market Risk Prediction

Abstract

Given that the financial market is influenced by political, economic and social factors, the instability it exhibits leads to an increase in investment risks, and traditional models have certain limitations in predicting financial market risks. Therefore, based on the RNN model and LSTM model, this paper reviews the application of the Transformer model in financial market risk prediction. By utilizing the attention mechanism of the Transformer model to predict complex high-frequency time series, and summarizing the fields where the Transformer model can be applied, a comparative analysis of the usage of each model was conducted, and the limitations of the Transformer model were also proposed. The study found that the Transformer model is more suitable for analyzing high-dynamic data and industries compared to traditional models. This review summarizes the application of the Transformer model in financial market risk prediction and adopts data from Wanfang Database and the National Philosophy and Social Sciences Literature Center. Relevant literature from 2018 to 2025 was collected, and when screening, articles with higher impact factors from journals and international conference papers were given priority. The topics of the literature include the attention mechanism of the Transformer model, risk response measures in the financial market, the prediction accuracy of the model, the field of natural language processing (NLP), and the spread trading market.