One of the major challenges in fire investigation is the determination of the cause of fire. The fire can be accidental or intentional. The determination of ignitable liquid residue (ILR) from fire debris helps the process and this process is called fire debris analysis in forensic science. This is one of the most complex areas in the field of forensics because of the evaporation of the ILR from the debris and the interference of the substrate matrix with the ILR if present. In the present, the final decisions in fire debris analysis are based on categorical statements and it only represents the qualitative but not the quantitative value of the data. The likelihood ratio approach is one of the most widely used methods in forensic science in expressing the evidentiary value. The purpose of this research is to introduce the likelihood ratios calculated by the Naïve Bayes approach. The data for this work was obtained by the Substrate and ILRC Databases from the National Center for Forensic Science. This project also contributed to the expansion of the Substrate Database by adding 1500 new substrate burn data records. The compounds identified from ignitable liquids and substrates were used to calculate the frequency of occurrences of the compounds in substrates and ignitable liquids. The presence or absence of the compounds was determined by the probabilities calculated by logistic regression. These frequencies of occurrences were used in the calculation of Naïve Bayes log likelihood ratios. The application, performance and validation of these models are discussed in this dissertation. These calculated log-likelihood ratios indicated that this method provides high evidentiary values in the classification of fire debris as positive for ILR in most cases but provided low evidentiary values in some other instances.
Identifer | oai:union.ndltd.org:ucf.edu/oai:stars.library.ucf.edu:etd-7838 |
Date | 01 January 2019 |
Creators | Akmeemana, Anuradha |
Publisher | University of Central Florida |
Source Sets | University of Central Florida |
Language | English |
Detected Language | English |
Type | text |
Format | application/pdf |
Source | Electronic Theses and Dissertations |
Page generated in 0.0019 seconds