Statistical matching rules and applications for common source identification

Abstract

We propose several statistical matching rules to address common source identification problems in forensic science, where a key question is determining whether two items originated from the same source. A statistical matching rule denotes a statistical decision-making method for source identification problems. In Chapter 2, we derive a parametric matching rule that minimizes a weighted sum of error probabilities under known density functions in the closed-set framework. In Chapter 3, we introduce a method to automatically align two similar footwear impressions and to identify the common source. In Chapter 4, we propose a nonparametric matching rule that leverages statistical classification. We showcase the performance of our proposed methods through numerical simulations or on various datasets in comparison to existing ones.