The invention discloses a patient feature depiction method and a false hospitalization behavior detection method based on the patient feature depiction method to solve the problem of insurance fraud behavior of false hospitalization in medical insurance. By combining statistics with machine learning, a hospitalization behavior of a patient can be effectively depicted so as to reduce artificial determining workload. The patient feature depiction method comprises the depictions for patient hospitalization features and patient basic personal information. The false hospitalization behavior detection method constructs a false hospitalization behavior model, which is constructed by the four procedures of feature depiction, cluster analysis, artificial screening and machine learning for model-setting. The false hospitalization behavior detection method effectively screens a false hospitalization behavior by utilizing the combination of the statistics and the machine learning, performs the feature depiction of the hospitalization behavior of the patient, effectively depicts the patient, screens out outliers by utilizing a method of clustering, and clustering can greatly reduce normal data so that the workload at an artificial screening stage is reduced; after the false hospitalization behavior model is set up, the hospitalization behavior of the patient is predicted by utilizing the same.