IDENTIFYING RISK CONDITIONS TO CRASH SEVERITY: GENETIC MINING RULE APPROACH

Abstract

This paper proposes a two-stage analytical framework to identify the critical risk conditions contributing to crash severity. The first stage develops a genetic mining rule (GMR) model to identify possible risk conditions best elucidating the degree of severity. The second stage then uses the mined risk conditions as dummy explanatory variables to formulate an ordered Probit model. The proposed two-stage analytical framework is applied to analyze the Taiwan's empirical one-vehicle crash data. A total of 38 rules are mined, which can achieve overall prediction rates of 75.10% in training and 73.80% in validation. Based on the results, six most critical risk conditions are identified and the corresponding countermeasures are addressed.