Finding criminal suspects by improving the accuracy of similarity measurement

Clustering technique was introduced to the field of crime data analysis for finding suspects, but traditional clustering methods used in existing application systems do not provide enough accuracy to meet the high requirements of police work. To solve the problem of low accuracy, we propose a hybrid similarity measurement, i.e., Segmented Multiple-Metric Similarity Measurement (SMMSM). In our method, compensation relationships among attributes are analyzed, attributes are grouped into multiple subsets, different measurements can be used in the meantime to measure the similarity of two objects, and the principles of classifying attributes are discussed. Experiment results show that our method has higher performance on accuracy and efficiency than traditional clustering methods.