Behaviour mining is an important issue in data mining .Peculiarity groups mining refers to finding the groups formed by those objects with similarity or common features from a dataset in which most of the data objects do not have the similar behaviours , such groups are called the peculiarity group .Therefore the peculiarity group has the feature of limit-objects and high-aggregation .In actual operation of medical insurance fund , a kind of seeking-medical-care fraudulent conduct called as “concerted fraudulent insurance action” shows the feature of peculiarity group .Conventional means as clustering or so are not suitable for the classification of peculiarity groups .By creating the adjacent peculiarity graph of the data objects and carrying out sparse treatment on it , we correspond the peculiarity groups search to the search of maximum complete sub-graph of adjacent peculiarity graph , and give a graph-based peculiarity groups mining algorithm .%行为挖掘是数据挖掘中一个重要的问题。特异群组挖掘是指在一个大部分数据对象不具有相似行为的数据集中，发现那些具有相似性或共同特征的对象形成的群组，称之为特异群组，因此特异群组具有的少对象、高聚集的特征。在医保基金实际运营中，一种称之为“一致骗保行为”的就医欺诈行为就表现出特异群组的特征。常规的聚类等方式不适用于对特异群组的分类，通过构建数据对象的特异邻接图并对其进行稀疏化处理，将特异群组搜索对应到特异邻接图的最大完全子图搜索上，给出一种基于图的特异群组挖掘算法。
展开▼
