Citation

Lamb DS, Lee C. Transp. Res. Rec. 2019; 2673(12): 234-243.

Copyright

(Copyright © 2019, Transportation Research Board, National Research Council, National Academy of Sciences USA, Publisher SAGE Publishing)

DOI

10.1177/0361198119848703

PMID

unavailable

Abstract

Motorcycle safety is usually included in the emphasis areas of Strategic Highway Safety Plans because of the over-representation in fatal traffic crashes relative to the proportion of motorcycle registrations. Understanding the trends and patterns of motorcycle fatalities requires unique considerations because, unlike passenger vehicles, trip purpose causes their riding patterns to be irregular. Most studies of motorcycle fatalities focus on trends at the national or state levels; however, concentrating on changes at the county level allows for more targeted intervention strategies. As this study shows, state patterns can be quite variable, with a small number of counties accounting for most motorcycle fatalities. This paper seeks to identify spatial and temporal patterns in total motorcycle fatalities at multiple scales (state and county). This research focuses on 3-year aggregates of 2004-2006 pre-recession and 2014-2016 post-recession total motorcycle fatalities from the Fatality Analysis Reporting System. To examine spatial patterns of change in total motorcycle fatalities, a stochastic gradient boosted random forest model from 2004 to 2006 was trained to predict the aggregate values in 2014 to 2016. The model was trained on variables that encompass economic, demographic, and climate factors. The results demonstrate the disaggregated variability in total fatalities. This approach identified several counties that had a significantly higher-than-expected increase in the number of fatalities that population increases cannot explain. Also, the approach shows potential as a method of setting targets for Safety Performance Management.

Language: en