Hybrid outlier mining algorithm based evaluation of client moral risk in insurance company
Wang Xiaoyun, Liu Danyue · 2010
Client moral risk in insurance industry arouses many problems such as insurance fraud, high loss ratio and adverse selection. Outlier detection which helps identify inconsistent records from large amounts of information data taken from policyholders, is becoming an important task of insurance companies. Data mining algorithm, which aims to identifying outliers, is acknowledged as a viable solution to discover clients with high moral risk. This paper presents a new algorithm combining the RB algorithm and density factor. It has higher precision and dose not need input parameters. Experiments are conducted using real life dataset from large insurance company. Comparison with RB algorithm through the experiments results reflects that the proposed algorithm is more effective and can serve as a detector of potentially inconsistent records.