Frequent Pattern Discovery Of Incremental Abnormal Medical Aggregation Behavior And Its Implementation On Spark

With the continuous development of the economy and society,people's quality of life is continuously improving,and our country's social security system is continuously improving.When people encounter medical difficulties,the healthcare fund will play an increasingly important role.The healthcare fund has providen strengthful safeguard for the social stability and the people's healthy life.However,outlaws are going to target people's life-saving money,using loopholes or flaws in the healthcare system,taking out healthcare funds,squandering the country's resources,causing great medical inequities and detrimental to the stability of the society.The illegal people fraudulently obtain healthcare fund in various forms,such as dismantling hospitalization,repeated medical treatment,and keep back the health insurance card.Currently,the common anti-fraud measures mainly rely on the experience of inspectors and the use of artificial means.In the case of a large amount of data,it is obviously not appropriate.Therefore,it is imperative to make use of technical means to identify fraudulent practices.This paper focuses on one of the healthcare defrauds which is the behavior of abnormal aggregation of patients,mining patients' behaviors by data mining algorithm,then deeply analyzes patients and hospitals based on the results of mining.On this basis,a fraud detection system is constructed.The main content is as follows:1.Healthcare data processing.Review the current problems in the patient's data,and perform data processing,including missing value processing,duplicate value processing,etc.,and carry out portraits of patients and hospitals to prepare data for in-depth analysis,besides lay the foundation for the construction of systems and other models.2.An incremental frequent pattern mining algorithm EFUFP is proposed based on support counts.An EFUFP algorithm for incremental frequent patterns based on support counts was designed to solve the problem of data batch update and the fraud scenario in which the behavior of abnormal aggregation of patients,and introduced the Spark programming model to implement the EFUFP algorithm on Spark platform to achieve rapid and efficient mining of frequent patterns under large-scale data.With the same result and similar spatial complexity,the time efficiency of EFUFP is increased by more than 10.7% compared with FUFP on local test,and is more than 26.8% higher than the FP-GROWTH algorithm on Spark platform.3.Construct abnormal medical behavior monitoring system and the proposed method of patient number anomaly detection in hospitals based on similarity analysis of time series.We use monitoring system to achieve the discovery,analysis and results display of the behavior of abnormal aggregation of patient,includes the analysis of the patient,and the analysis of the hospital.The analysis of patient is mainly based on the use of rules to identify the patient's illness,costs,and other abnormalities.For the hospital analysis,an abnormal patient number detection method in the hospital based on similarity analysis of time series was proposed,and the abnormal number of patient in the hospital was detected dynamically.