| Medicare as a part of the entire social security system plays a great role in health insurance of all the labors.However,the frauds and violations caused by the nonsymmetrical information come into being frequently,meanwhile,the means of frauds becomes more hidden and professional,resulting into lots of difficulties in risk management.Under this background,introducing the Agent technologies into the medicare domain to conduct real-time detection of medicare frauds and violations is important for medicare supervision and fund loss decrease.Policyholders to see a doctor are in a open dynamic surrounding,during the process,doctors,policyholders and insurance companies are intelligent and rational agents and have the abilities to related problems solving,in addition,they can predict the consequences of their own behaviors.In a word,all of the above mentioned matches with Agent's features.Therefore,a medical fraud detection system based on multi-Agent is proposed in this dissertation.In the system,the medical institutes and insurance companies are taken as the corresponding Agents,frauds detection as the process of multi-Agent decision making.The blackboard among these Agents provides a public workspace to obtain information and intention from each Agents, the fraud detection actions are so identified through making use of Agent's intelligence and multi-Agent coordination.The main contents are described as follows:1,Analyzes the currently used abnormality detection methods and Agent related theories and technologies,applications of Agent technologies to abnormality detection is also explored.2,On the basis of analysis on the medicare domain problems,the subject and flow of abnormality detection are decided.Defines the medical actions to be detected according to analyze the medical data.3,In terms of features of Agent technologies,shows a model of medical fraud detection system based on multi-Agent,discusses its working mechanism and designs its architecture structure and inner functional Agents.4,Implements the detection system prototype on the platform JADE. Furthermore,aiming at the effect-effect similarity in outpatient medication gives the abnormality detection strategies and verifies the system feasibility via test on a real-world data. |