Article Preview
TopIntroduction
The collapse of the Bretton Woods system has created a lasting influence on international financial markets that have become extremely volatile under accelerating economic globalization (Nikulin & Pekhterev, 2021). In such a financial environment, enterprises, financial institutions, and individual investors might have to bear various unprecedented risks (Nosan, 2019), which seriously harms the healthy development of the national economy and global financial markets (Olivier & Lieven, 2019). In particular, the 2008 subprime mortgage crisis has featured the bankruptcy of numerous enterprises and huge losses of financial institutions (Soares et al., 2021). Worse still, without a global financial regulation system, the international financial market is becoming ever more complicated (Thomas, 2017). Generally, enterprises, financial institutions, and individual investors utilize effective risk prediction models to analyze financial data and avert risks (Ouyang et al., 2021). Compared with manual analysis, model prediction is more objective and can interpret financial data from multiple angles, avoiding potential systematic risks (Zhou, 2017). Therefore, the study of the financial market risk management system plays an important role in ensuring the stability of the national financial market.
Initially, the variance-based risk management system is used for the financial market, which considers the return surplus over the average into risk management and takes the average return as the benchmark (Sun, 2021); although the average deviation is considered in the model, it is not suitable to predict small probability events (Ma et al., 2020). Later, the Value at Risk (VaR) model comes into being, which can effectively control the portfolio risks under the same number of securities (Elena, 2019); still, this model is unstable and highly dependent on product types (Hee & Christian, 2021). Afterward, the Coherentmeasure of Risk (COR) model is proposed, which has shown great practical significance in the determination of capital and portfolio of banks (Veryzhenko, 2021); yet, this model has presented low-linearization ability because of high dimensionality and complex calculation (Dolfin, 2019). Recently, a new risk management model is proposed based on deep learning. Deep learning learns the inherent laws and representation features of sample data and uses them to interpret text, image, and sound data. Deep learning aims to intellectualize machines with the same analytical and learning ability as humans to recognize various data (Pang et al., 2019), which is a complex machine learning algorithm and is extremely effective in speech and image recognition. At present, the risk management model based on deep learning has shown many advantages, such as powerful calculation, good adaptability, strong learning ability, and wide coverage, as well as high prediction accuracy and wide application. Although the application of deep learning algorithms in financial risk management has been matured, there are still many problems (Valeriane & Wolfgang, 2020). Therefore, the study of financial credit risk management systems based on deep learning has important research value for promoting the development of the financial industry.