Research on Bank Anti-Fraud Model Based on K-Means and Hidden Markov Model

Abstract

Internet finance is developing rapidly. As online payments such as Alipay and WeChat Pay become more and more popular, cases of fraud associated with are also rising. In this paper, we describe the entire process of fraud detection using Hidden Markov model (HMM). We use the k-means algorithm to symbolize the transaction amount and frequency sequence of a bank account. This sequence is used to build and test the model. An HMM is initially trained with the normal behavior of an account. If an incoming credit card transaction is not accepted by the trained HMM with sufficiently high probability, it is considered to be fraudulent. We illustrate the feasibility of the model through simulation experiments and verify the validity of the model with real-world bank transaction data. Especially, in the case of enough historical transactions, this method performs well for low, medium frequency and amount of user groups.