Insurance fraud identification research based on fuzzy support vector machine with dual membership

Insurance frauds continue to emerge at home and abroad, the insurance company not only faced with the failure to identify fraud which lead to abuse lose a threat, but also bear the loss of recognizing true insurance claims as insurance fraud. For “overlap” problem in insurance fraud samples, this paper constructs the fuzzy support vector machine model with dual membership, which assigns each insurance fraud sample with dual membership by its relativity to the distance of the two types of sample mean vector. The dual membership can characterize the probability of each insurance fraud sample belonging to two categories. The empirical experiments indicate that the result of dual membership fuzzy support vector machine model is better than the existing insurance fraud recognition model.