The Detection Model Of Normal And Abnormal Instances Based-on Process-Mining

Now there are lots of studies on detecting of abnormal cases. It is mostly studied how to build distinguish standard or feature library artificially. This kind of method not only make preparations troublesome, but later discriminating ability is limited and low accuracy and has obvious limitations in many applications.Machine recognition is an important field of artificial intelligence, using widely in industrial production, medical, finance etc. in recent years. But the prophase work is still rely on setting standards by experts, this standard due to can't get rid of subjectivity, and makes the discernment accuracy is reduced greatly.This paper firstly introduces the normal and abnormal cases related to discern the research background and the recent research status,secondly introduces the basic knowledge relevant to the process mining and data mining. Aiming at the existing method has its own deficiencies and summing up predecessors'experience and research basis, this paper proposed a model (the normal and abnormal detection model of normal and abnormal instances, DMNAI) based on process-mining, throughing frequent patterns to extract the case feature, using neural network classifier to detect cases. Consequently,we can void a manually set standards subjectively.This paper based on the framework of research and compared the DMNAI model to the DHFA model. In the DHFA framework, because of the choice of characteristics is mechanical work, making query this model generalization ability are relatively weak. DMNAI model through process mining to extract features, establish a more widely detection model.Experiments show that DMNAI model after the experimental data validation, can effectively discriminating abnormal cases.In the end, with online shopping flow as the example, it is proved the feasibility and the accuracuy of the model.