Abstract: With the advancement of China’s reform and opening up in the 1980s, the emergence of joint-stock companies has led to widespread attention on the construction of the stock market. As more and more joint-stock companies appeared and stocks were issued, the stock exchange emerged to deepen the reform and opening up of the financial industry. During the period from 1990 to 1991, the Shanghai Stock Exchange and the Shenzhen Stock Exchange were established, which opened the construction of China’s stock market. Although the Chinese stock market started later, it has been continuously improved through the advancement of reform. As an important component of the financial market, the stock market not only plays a crucial role for investors and listed companies but also stabilizes the national financial order and improves the ability to withstand risks. Throughout history, many financial crises have been caused by stock market crashes, such as the Great Crash of 1929 in the US, the 1990 Japanese stock market collapse, the 2015 Chinese stock market plunge, and the global stock market crash in 2020 caused by the COVID-19 pandemic. Studies have also shown that the stock market is the largest risk output and receiver, and a stock market crash can cause panic among investors, lead to financial crises for listed companies, and even affect the overall operation of the socio-economy. Therefore, it is essential to be able to alert and improve the ability to withstand risks for investors, listed companies, and the government.
In the big data era, traditional linear prediction methods such as data simplification, composite index method, and financial stress index method are no longer accurate in describing financial market risks. The emergence of new intelligent machine learning algorithms such as Decision Trees (DT), Logistic Regression (LR), Random Forests (RF), Artificial Neural Networks (ANN), Copula, and Support Vector Machine (SVM) have greatly addressed the problems of the big data era. As a machine learning method, SVM is frequently used for data analysis and regression problems due to its strong non-linear fitting ability, simple learning rules, easy implementation by computers, and the ability to achieve optimal decisions using a small number of support vectors. It effectively solves the complexity of indicators in the era of big data. However, traditional SVM is sensitive to missing data, and the selection of penalty coefficient C and kernel function parameter g is subjective and empirical, which can consume a large amount of memory and time in the case of large samples. From existing research, SVM has been widely used in company financial risk warning and financial market warning, and has achieved certain research results. However, it has been less applied in stock market risk warning. The key to preventing risks lies in constructing a reasonable early warning model, so the adaptability of the SVM model to stock market prediction is a subject that needs further research.
Given the importance of risk warning for the stock market, this article proposes a Grey Wolf Optimizer Support Vector Machine (GWO-SVM) stock market risk warning model to improve China’s stock market risk warning ability. This is in response to traditional SVM problems, such as difficulties in parameter selection and low prediction accuracy. The effectiveness of the model was tested using the Mean Absolute Error (MAE) and Mean Squared Error (MSE). The Grey Wolf Optimization algorithm is an intelligent optimization algorithm proposed by scholars from Griffith University in Australia in 2014. Inspired by the hunting behavior of grey wolves, this algorithm is a type of optimization search method that has strong convergence performance, fewer parameters, and is easy to implement. It can significantly improve the efficiency and prediction accuracy of SVM. Our paper focuses on the daily returns and volatility of eight major stock market indices in China. Daily returns can comprehensively reflect the price changes and trends of stocks, while volatility can effectively measure market sentiment and help managers judge the macro trend of the market. Moreover, these eight indices have a broad coverage and can basically represent the operation of the entire stock market. They are often used as benchmark indices to measure the overall market risk. Therefore, daily rate of return and volatility are selected as the research objects, with data collected from CSMAR and RESSET databases. The research results show that compared to SVM, GS-SVM, GA-SVM, and PSO-SVM, GWO-SVM has an average runtime efficiency that is 330% longer than the other three optimization algorithms. Meanwhile, the GWO-SVM model has an average decrease of 4% in MAE and 5% in MSE in predicting daily returns, and it also shows a highly fitting trend in predicting daily volatility. Therefore, the model can effectively improve the accuracy and efficiency of predicting stock market risks.By comparing the original and predicted data, GWO-SVM can accurately predict the fluctuation of the stock index, providing new ideas for stock market risk prediction in China. Future research will focus on characterizing risk indices and further optimizing the model to better analyze and predict stock market risks.

Key words: stock market risk, grey wolf optimization algorithm, support vector machine, risk warning

摘要： 鉴于预警股票市场风险的重要性,为提高我国股票市场风险的预警能力,针对传统支持向量机(SVM)参数选择困难和预测精度不高等问题,基于灰狼优化算法(GWO)提出灰狼算法支持向量机(GWO-SVM)股票市场风险预警模型,并利用平均绝对误差(MAE)和均方误差(MSE)检验了有效性。研究结果表明,与SVM、GS-SVM、GA-SVM、PSO-SVM相比, GWO-SVM模型对日收益率预测的MAE平均降低了4%,MSE平均降低了5%,能有效提高股票市场风险的预测精度和效率。通过原始-预测数据的对比,GWO-SVM能较为准确地预测出股票指数的波动情况,为我国股票市场风险预测提供了新的思路。

关键词: 股票市场风险, 灰狼优化算法, 支持向量机, 风险预警