With the economic development, more and more people participate in medical insurance and enjoy the benefits of medical insurance. However, medical insurance fraud has brought tremendous losses to the medical insurance fund. This paper presents a two-stage outlier detection method for medical insurance fraud detection. In the first stage, weighted k-means algorithm is used to cluster the data set and prune the result set, where the weight w is calculated by using the particle swarm optimization algorithm to minimize the evaluation function of the weight index. The second stage adopts the improv...