One of the most effective approaches interpreting accident causality is from the chain perspective. However, this approach has been applied only to measuring safety improvement for specific countermeasures. This paper emphasizes that the chain concept, which reflects the generating process of an accident, should be borne in mind when analyzing accidents and preventions. A safety framework of the driving behavior is constructed from the chain perspective in support of this argument. Also, a two-stage approach is proposed to adopt the idea in analyzing accident data. The approach includes applying classification techniques at the first stage and causal inference models at the second stage. The relevant methodologies are introduced and possible issues are discussed. It is believed that the proposed framework and approach would help further studies to analyze and interpret accident data in a more thorough perspective. (Author/publisher)
Abstract