Bayes factors were used to determine the strength of evidence that two crimes have been committed by the same person. Using concepts from agglomerative hierarchical clustering, the Bayes factors for crime pairs were combined to provide similarity measures in comparing two crime series. This facilitates crime-series clustering and identification, as well as suspect prioritization. The models' ability to make correct linkages and predictions was demonstrated under real-world scenarios that involve a large number of solved and unsolved breaking-and-entering crimes. A naïve Bayes model for pairwise case linkage identified 82 percent of known linkages, with a 5-percent false positive rate. For crime-series identification, 77-89 percent of the additional crimes in a series were identified from a ranked list of 50 incidents. 6 tables, 6 figures, and 71 references
Similar Publications
- Long-Term Memory in Adults Exposed to Childhood Violence: Remembering Genital Contact Nearly 20 Years Later
- Artificial Neural Network System for Classification of Offenders in Murder and Rape Cases
- Factors Affecting the Accuracy of Eyewitness Identifications Investigating the Validity of the US Supreme Court's Guidelines - Final Report