BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Assessing Evidentiary Value in Fire Debris Analysis - Michael  Sig
 man  (University of Central Florida)
DTSTART:20161110T153000Z
DTEND:20161110T161500Z
UID:TALK68926@talks.cam.ac.uk
CONTACT:INI IT
DESCRIPTION:<span>Co-author: Mary R. Williams (National Center for  Forens
 ic Science\, University of Central Florida) <br></span> <span><br>This pre
 sentation will examine the calculation of a likelihood ratio to  assess th
 e evidentiary value of fire debris analysis results. Models based on  supp
 ort vector machine (SVM)\, linear and quadratic discriminant analysis (LDA
   and QDA) and k-nearest neighbors (kNN) methods were examined for binary 
  classification of fire debris samples as positive or negative for ignitab
 le  liquid residue (ILR). Computational mixing of data from ignitable liqu
 id and  substrate pyrolysis databases was used to generate training and cr
 oss validation  samples. A second validation was performed on fire debris 
 data from large-scale  research burns\, for which the ground truth (positi
 ve or negative for ILR) was  assigned by an analyst with access to the gas
  chromatography-mass spectrometry  data for the ignitable liquid used in t
 he burn. The probabilities of class  membership were calculated using an u
 ninformative prior and a likelihood ratio  was calculated from the resulti
 ng class membership probabilities . The SVM  method demonstrated a high di
 scrimination\, low error rate and good calibration  for the cross-validati
 on data\; however\, the performance decreased significantly  for the fire 
 debris validation data\, as indicated by a significant decrease in  the ar
 ea under the receiver operating characteristic (ROC) curve. The QDA and  k
 NN methods showed performance trends similar to those of SVM. The LDA meth
 od  gave poorer discrimination\, higher error rates and slightly poorer ca
 libration  for the cross validation data\; however the performance did not
  deteriorate for  the fire debris validation data.</span>
LOCATION:Seminar Room 1\, Newton Institute
END:VEVENT
END:VCALENDAR