论文部分内容阅读
本文探讨商业银行如何利用贝叶斯分类技术构建企业客户财务危机预测模型。本文使用财务比率作为评价企业绩效的特征属性,并考察两个不同的贝叶斯模型在估计企业客户发生财务危机的后验概率方面的有效性。一个比较简单但有较多的假设,即朴素贝叶斯模型;另一个某种程度上更为复杂但有更少的假设,即组合属性贝叶斯模型。研究发现,与朴素贝叶斯模型相比,由于组合属性贝叶斯模型更好地反映了变量之间潜在的联合分布,因此它能在历史数据支持下估计所要求的概率并做出更精确的预测。所提出的模型可以作为辅助银行审核者做出正确而快速决策的有用工具。
This article explores how commercial banks use Bayesian classification technology to build a corporate customer financial crisis prediction model. This paper uses the financial ratios as a characteristic attribute for evaluating firm performance and examines the effectiveness of two different Bayesian models in estimating the posterior probability of a financial crisis for corporate clients. A simpler but more presumptive one is the Naïve Bayesian model; another is a somewhat more complicated but less presumptive one, the Bayesian Model of Combination Attributes. The study found that, compared with the naive Bayesian model, the Bayesian model of combination attributes can estimate the required probability with the support of historical data and make more accurate because the Bayesian model of combined attributes better reflects the potential joint distribution among the variables Prediction. The proposed model can be a useful tool to assist bank reviewers in making the right and quick decisions.