基于Spark机器学习实现医疗保险关联频繁模式的欺诈行为挖掘技术探讨  

Research on Fraud Mining Technology to Realize Frequent Patterns of Medical Insurance Association Based on Spark Machine Learning

作　　者：刘鹏[1] LIU Peng

机构地区：[1]天津市第一中心医院

出　　处：《中国数字医学》2019年第5期15-18,共4页China Digital Medicine

基　　金：中华医学会医学教育分会,中国高等教育学会医学教育专业委员会2018年医学教育研究立项课题(编号:2018B-N05043)~~

摘　　要：医疗保险基金在社会稳定与人民健康生活水平提高中所承担的作用日益重要。基金审计所面对的是医疗机构几何级增长的“知识密集型”各类医学数据,传统的骗保行为甄别统计分析方法无法适用此类多样性与非结构化数据应用场景,必然导致探索以机器学习理论为基础的大数据分析方法的研究应用。围绕电子病历文本关联规则挖掘问题,探索利用Spark机器学习技术对关联规则挖掘算法在电子病历文本表达数据中的应用进行深入研究,在此基础上构建医疗保险欺诈行为审计系统的核心逻辑,用以深层次分析医疗机构和患者双维度挖掘效果。Medical insurance fund plays an increasingly important role in social stability and the improvement of people's health and living standards. The fund audit institute collects all kinds of knowledge-intensive medical data with the geometric progression growth of medical institutions. The traditional statistical analysis methods of fraud screening cannot be applied to such diverse and unstructured data application scenarios, which inevitably leads to the exploration of the research and application of big data analysis methods based on machine learning theory. Focusing on the problem of mining association rules in electronic medical record text, this paper explores the application of Spark machine learning technology in the mining algorithm of association rules and its parallelization, as well as the mining of association rules in the expression data of electronic medical record text. On this basis, the core logic of medical insurance fraud auditing system is constructed to deeply analyze medical institutions and patients and two-dimensional mining effect.

关 键 词：大数据分析 电子病历 SPARK 机器学习 频繁模式 数据挖掘 

分 类 号：R319[医药卫生—基础医学] TP302.1[自动化与计算机技术—计算机系统结构]