An Entity Resolution Model based on Time-Sensitive for Telecom Fraud Crime Investigation

Abstract

Entity resolution is an important part of the social network analysis, and it aims to identify the real corresponding entity out of a confusion of entities. Based on the standard terminology in graph entity recognition, this paper proposed an improved Jacquard similarity measure approach according to features of telecom fraud crime, and presented a novel entity resolution algorithm which utilizes some new measures as time step and time-sensitive coefficient. This algorithm can be applied to multi-time-step conditions in order to analyze the entity’s changes. By taking into full account of the movement area of the suspects established by the Police Bureau, it excludes the impossible telecom entities, and eventually discovers the result cluster of the given mobile numbers that possibly belong to the same entity. This approach can help the Police Bureau narrow the investigation scope and maximize the anti-crime effects in the information age. Based on the phone-call data collected by the police bureau, this algorithm that has been applied in the telecom frauds well show its feasibility, practicality and effectiveness.