Traffic Safety Prediction
Application of ML algorithms in the context of traffic safety analysis
Overview
Three algorithms are tested for crash severity and frequency analysis, including decision tree (CART), random forest, and model-based recursive partitioning (MOB). Model development, optimization, and evaluation are jointly compared with statistical approaches in the study field (i.e., logistic regression and negative binomial regression models).
Data Source
- Pennsylvania Crash Information Tool
- PennDOT's GIS Open Data Portal
- PennDOT's VideoLog
- Research projects
- Google Earth, etc.