This study use data mining to detect assessment issues for hospitals with abnormal charge onhealth care insurance. We select dispensaries of Chinese medicine, western medicine, and dentistmainly located at Northern Taiwan as the studying objects, and use data mining to detect theirrequested outpatient charges to see is there any falseness existed. We explore data mining byusing neural network, discriminant analysis, and logistic regression to evaluate and recognize theabnormal cases. Comparing the results assessing by different methods, we find that neuralnetwork model can monitor hospitals of Chinese medicine, western medicine while logisticregression is more suitable for dentist. This prediction model can assist the auditors to check thepossible abnormal hospitals in order to improve the overflow and untrue charge on healthy careinsurance. This model can also be extended to different areas, professions, or be advanced formore specified data mining model.