The Method of Naive Bayesian Network Clustering for Industry Risk Prediction

摘要：

　　At present, there are not effective and practical methods for prediction with small or incomplete examples.The naive Bayesian network is combined with Gibbs sampling in static and dynamic clustering prediction for this situation.For improving the reliability of prediction, Gaussian kernel function is applied in estimating conditional density.And the methods of ensemble static and dynamic clustering are developed by introducing and configuring shape parameter for reducing error produced by single clustering.Enterprise risk data and international standard classification data are engaged in experiment and analysis about ensemble clustering.Experimental results show that ensemble clustering of naive Bayesian network is more effective and more reliable.

关键词： risk prediction small example sets static clustering dynamic clustering ensemble clustering

作者: Wang Shuangcheng Shao Jun Leng Cuiping

作者单位: The School of Mathematics and Information, Shanghai Lixin University of Commerce The School of Accounting and Finance, Shanghai Lixin University of Commerce

会议类型: 国际会议

会议名称: The 4th Conference on Chinas Economic Operation Risk Management(2010·Shanghai)(第四届中国立信风险管理论坛)

会议地点: 上海

会议语种:英文

页码: 172-180

在线出版日期: 2010-10-14（万方平台首次上网日期，不代表论文的发表时间）