Relationship Networks for Inverse Graph Partitioning

We describe multiple relationship networks constructed from data in four detention centers. The personal information, social information, and case information of all detainees in these detention centers are collected and transformed through data protection measures, such as anonymization and encryption. Three kinds of relationships between detainees are extracted from their information, including the relationship of joint crime, the relationship of fellow-townsman, and the relationship of the same type of crime. Hence, we generate multiple relationship networks, including single-layer unweighted relationship networks, multi-layer unweighted relationship networks, and multi-layer weighted relationship networks. We expect the open of this dataset will allow researchers to advance on the study of graph optimization problems, such as inverse graph partitioning and anti-community detection.