This article discusses the possibility of using data-driven prediction models to enhance fire origin determination; reporting on a proof-of-concept study that simulated 600 fire scenarios, the paper discusses the data collection, testing and evaluation, and findings.
This study evaluates data-driven methods for fire origin determination in pre-flashover fires based on interior fire pattern distribution. A proof-of-concept study was conducted by simulating six hundred compartment fire scenarios to generate data for model development. Heat exposure of interior surfaces was recorded to identify potential areas with fire patterns. Models such as Partial Least Squares Regression, Random Forest, and Support Vector Regression were tested, with Random Forest showing the best performance in locating fire origin. This study demonstrates the potential of data-driven models for identifying fire origins but highlights limitations in complex fire sites and scenarios. (Published Abstract Provided)
Downloads
Similar Publications
- Randomized Clinical Trial Pilot Study of Prolonged Exposure Versus Present Fcentred affect regulation therapy for PTSD and anger problems with male military combat veterans
- Correction: Shelly Y. Shih; et al.; Applications of Probe Capture Enrichment Next Generation Sequencing for Whole Mitochondrial Genome and 426 Nuclear SNPs for Forensically Challenging Samples. Genes 2018, 9, 49
- rASUDAS A New Web-Based Application for Estimating Ancestry from Tooth Morphology