An Outlier Analysis Of Doctors’ Medication Based On CASO Similarity And Chameleon Algorithm

In recent years, there are more and more reports about the relationship between doctors and patients. The most prominent problem are that doctors over check list, excessive use of drugs and big prescription, these problems also let people complain frequently. This situation is partly due to the doctor’s personal qualifications and other unconscious behavior, but a large part of the reason is that doctors intend to be. Driving by the existing medical interests mechanism, pharmaceutical profit can be withdraw cash through medical services. Doctors and medical representatives to reach a consensus, use the recommended drugs by medical representative instead of high cost of drugs, this may delay the diagnosis and treatment of the disease, increases the economic burden of the patient. Through the outlier analysis of medical treatment technology can be effective for doctors to automatically detect outlier medication and to help the relevant department to detect whether the doctors are rational drug use, and take the necessary regulatory measures.The purpose of the outlier analysis of the doctor’s medication is to test whether the prescription of a certain kind of disease is outlier. Firstly, in the data preprocessing stage, the prescription is the data object, and each drug category is marked as a property to the data. And then to make the qualitative analysis of the data, the data is divided into several groups according to the department and the name of the disease. For each similar group, the Coupled Attribute Similarity for Objects(CASO) is used to measure the similarity between different objects. The similarity calculation results using the chameleon algorithm to cluster the object set. Finally, the clustering results are mapped from the prescription to the doctors, outlier detection of the doctor’s medication was carried out by using the ratio of the prescribed outliers(PO) and the ratio of the cluster prescriptions(CP).In similarity calculation stage, this thesis calculates similarity in the real data using the Jaccard similarity coefficient calculation method and CASO similarity degree. Experiment results show that the experimental results of CASO is better than Jaccard, but the computational efficiency is less than Jaccard. The experimental results on the real data sets show that the method has achieved good results.