This paper presents the results of an analysis focusing on the recognition of large truck-involved work zone crash patterns with fatality outcomes. This study applied random oversampling and systematic oversampling techniques using a ten-year large truck involved work zone crash data in the state of Florida. Decision trees and random forest models were consequently built for the raw and resampled datasets. Results showed that the combination of oversampling with ensemble random forest technique significantly improved model performance in predicting fatality crashes. In addition, results showed that certain driver actions, combined with roadway conditions, proximity to signalized intersections, lighting conditions, and vehicle restraint equipment were significantly associated with fatality outcomes. Among different driver actions, improper turns, failing to keep proper lane, and wrong-way driving were critical movements. Among different types of trucks, trucks with trailers seem to be less likely to be involved in fatal crashes.

CICTP