Clinical Abnormal Behavior Detection Based On Sequential Pattern Mining

At present, medical fraud is common. And medical disputes are increasing day by day, so the relationship between doctors and patients is becoming increasingly tense. Because it is difficult to make a definition of medical fraud, regulators can not supervise medical institutions and medical workers well. Consequently, establishing a clinical abnormal behavior detection system for providing decision basis for the regulators to define the medical fraud, which is of great significance for the prevention and supervision of the irregularities and fraud in medical affairs and the relief of the strain on the doctor-patient relationship.Data mining technology has been widely used in various fields. According to the concept of clinical pathway, the characteristics of clinical medical behavior and medical fraud are analyzed. Then, it uses the weighted sequential pattern mining algorithm to get the pattern of clinical medical behavior, and researches the anomaly detection technology based on sequential pattern mining. Finally, it constructs a clinical abnormal behavior detection model to discover the abnormal clinical medical behavior.The research work in this article includes:(1)The weighted sequential pattern mining PT-WPSN algorithm is researched. The PT-WPSN algorithm uses a modified weighted prefix sequence tree storage structure, combined with the theory of pattern growth. At the same time, introducing the weight into the mining process, and finally, mining the clinical weighted sequence pattern.(2)Analyzing the event sequence similarity matching algorithm, then, propose the clinical sequence similarity matching method which is based on the sequence edit distance. In the calculation of the sequence edit distance, the similarity comparison of medical behavior with the effect, dosage and price attribute is added, and using the dynamic programming theory to get the minimum sequence edit distance, then determine whether the clinical sequence is similar.(3)Constructing the clinical abnormal behavior detection model. Based on the theory of using the degree that detection behavior patterns deviates from normal behavior patterns as the basis of anomaly judgement, constructing the model by combining clinical behavior patterns mining, pattern comparing, sequence similarity matching and scoring mechanism.(4)Implementing the prototype system. Designing the architecture of the prototype system and coding it, and finally, by evaluating and analyzing the ability of the system to identify the abnormal behavior to verify the correctness and usability of the model.