AK-Modes: A weighted clustering algorithm for finding similar case subsets
Lianhang Ma, Yefang Chen, Hao Huang · 2010
Finding similar crime case subsets is an important task for intelligence analysts in crime investigation. It can not only provide multiple clues to solve crimes but also improve efficiency to catch the criminals. However, the conventional approach by querying specific attributes in relational databases has two defects: first, it is relatively of poor efficiency when a lot of incidents have to be handled; second, the querying process can not reflect the importance of attributes in different case categories. In this paper, we propose a two-phase clustering algorithm called AK-Modes to automatically find the similar case subsets from large datasets. In the attribute-weighing phase, we compute the weight of each attribute related to an offender's behavior trait using the concept Information Gain Ratio (IGR) in classification domain. Then the result of attribute-weighing phase is utilized in the clustering process to find the similar case subsets. Experiments show that AK-Modes is effective and can find significant results.