Relating Freeway Traffic Accidents to Inductive Loop Detector Data using Logistic Regression

Abstract

This study proposes a methodology for utilizing loop detector data to understand the mechanisms of accident occurrence on freeways. Loop detector data were processed to derive useful traffic parameters representing spatio-temporal variations of traffic conditions, in order to contextualize which lead to accident events. Binary logistic regression modeling was conducted to correlate traffic variables with the accident data that were collected from the Seohaean freeway during the three years, from 2004 to 2006. Outlier filtering and data imputation were also performed to prepare for more reliable traffic variables. The outputs of the model, which are actually probabilistic measures of accident occurrence, can be utilized not only in designing warning information systems but also in evaluating the effectiveness of various traffic operation strategies in terms of traffic safety.