Identity Matching Based on Probabilistic Relational Models

Identity management is critical to various organizational practices ranging from citizen services to crime investigation. The task of searching for a specific identity is difficult because multiple identity representations may exist due to issues related to unintentional errors and intentional deception. In this study we propose a probabilistic relational model (PRM) based approach to match identities in databases. By exploring a database relational structure, we derive three categories of features, namely personal identity features, social activity features, and social relationship features. Based on these derived features, a probabilistic prediction model can be constructed to make a matching decision on a pair of identities. An experimental study using a real criminal dataset demonstrates the effectiveness of the proposed PRM-based approach. By incorporating social activity features, the average precision of identity matching increased from 53.73 % to 54.64%; furthermore, the incorporation of social relation features increased the average precision to 68.27%.