Graduate Theses, Dissertations, and Problem Reports

Fall

2018

Thesis

MS

College

Eberly College of Arts and Sciences

Department

Forensic and Investigative Science

Jacqueline Speir

E. James Harner

E. James Harner

Lucas Mentch

Abstract

The reproducibility of experienced-based forensic pattern interpretation is founded on the notion that domain-specific knowledge can be successfully distributed and applied among experts within a group. This assumption persists, even when the examination is complicated by variations in case circumstances, such as impression clarity and totality, as well as media, substrate, collection mechanism and enhancement. While it is further theorized that many of these factors (as well as additional confounding factors) are at play during an examination, the manner and extent to which these sources of variability affect the examination of footwear evidence remain unclear. In order to explore this hypothesis, a data mining technique called dominance-based rough set approach (DRSA) was applied to characterize the novice examiners’ decision-making process, due to its ability to capture useful information from a set of hybrid data with latent preference orders and discover knowledge in the form of decision rules. Through this approach, two objectives were addressed: the identification of factors that affect footwear examination and conclusions within the novice group, and the evaluation of decision rule quality as a function of support, strength, certainty and lift factors.

The results of the study showed that in general, novice examiners’ case assessments were found to be outside the acceptable conclusion range more than 50\% of the time, with general tendencies to assign ambiguous conclusions, such as limited association of class characteristics" and lacks sufficient detail," rather than more definitive ones such as identification" or exclusion." When assessments were further explored using DRSA, 23 decision rules were induced (13 \textit{certain} and 10 \textit{possible}). Of the 13 \textit{certain} rules, 75\% of the induced rules were dominated by the examiner’s background, rather than case attributes, and 50\% of the \textit{possible} rules indicated that media type was a prevalent factor in the examiners’ determination of similarity/dissimilarity, as they attempted to interpret media-substrate interaction and reconcile this interpretation with SWGTREAD conclusion guidelines. Only when examiner attributes were excluded from the analysis, forcing the induction of rules based on case attributes only, did case-based features become prominent, but only with very low rule-support. In the second phase of work related to this project, the nature and type of rules induced based on expert assessments will be examined and compared to those generated from this novice set in order to compare and interpret the manner in which domain-specific knowledge dominates induced rules.

Embargo Reason

Publication Pending

COinS