基于一般聚类分析的中国金融系统性风险评估研究

Research on China’s Financial Systematic Risk Assessment Based on General Clustering Analysis

通过选取我国2007年1月到2016年12月期间影响中国金融系统稳定的经济数据,将所选变量归为8个维度:金融机构、货币市场、房地产、股票市场、债券市场、政府部门、外汇市场、期货市场。利用熵值法测度我国金融系统性风险综合指标,并通过灰色关联判断不同时间段中,不同市场维度对金融系统性风险影响的重要程度。然后采用一般聚类算法,将我国的金融系统性风险状态分为轻风险、中风险和高风险,结果与我国实际情况基本相符合,所构建的金融系统性风险测度方法能在一定程度上识别风险。最后,使用支持向量机、随机森林、贝叶斯、决策树、BP神经网络,对我国面临的金融系统性风险进行分类预测,为我国金融系统性风险的预测和监管提供预防及监管信息。

By selecting China’s economic data that reflects China’s financial system between January 2007 and December 2016.The selected variables are classified into eight dimensions:financial institutions,money markets,real estate,stock markets,bond markets,government departments,foreign exchange markets,futures markets.Using the entropy method to measure the comprehensive index of financial system risk in China,and judge the importance of different market dimensions to the financial system risk through gray correlation judgment in different time periods.And then use the general clusteringalgorithm to divide our financial system risk state into light risk,risk and high risk.The result is consistent with the actual situation of our country.To a certain extent,the financial system risk measure method can identify risk.Finally,we use support vector machine,stochastic forest,Bayesian,decision tree and BP neural network to classify and forecast the financial system risk of our country,and provide prevention and supervision information for the forecast and supervision of financial system risk in China.