摘要
运用多元统计和概率神经网络的方法,结合实际研究项目,提出医疗保险中的核保风险分析模型.它是利用已知信息对投保人的患病风险以及属于何种风险类别进行分析和判别.其做法是首先对数据库用Kendall(tau-b)进行相关性检验,剔除导致疾病发生的相关性较小的因素,然后选取17 000条数据作为训练集,应用概率神经网络进行训练.训练完成后,选取另外1 000条数据作为测试集,检验预测结果.这种模型用MATLAB软件实现,具有可操作性,并可推广到相应的保险和金融等领域的风险分析问题中去.
Based on multivariate statistic and probabilistic neural network method associated with a practical project, a risk analysis model in medical insurance is presented. According to the available information, the risk of disease and the type of risk can be tested and analyzed. The Kendall（tau-b） method for calculation of correlation between the factors and the disease is first used to eliminate factors of little relation of the disease. Secondly, 17 000 records in the database are chosen as a training set to build a probabilistic neural network model. Finally, 1 000 records are chosen as a testing set to test accuracy of the model. The model is implemented using MATLAB, and it can be generalized and applied to insurance and financial regions.
出处
《上海大学学报（自然科学版）》
CAS
CSCD
北大核心
2006年第1期40-44,65,共6页
Journal of Shanghai University:Natural Science Edition
基金
上海市重点学科建设资助项目
