Classification Trees and Association Rules for Exploratory Analysis of Powered Two-Wheeler Crashes

Aim of the study was the exploratory analysis of powered two-wheelers crashes in Italy in order to: (1) detect interdependence among crash patterns as well as dissimilarities among patterns, that cannot be deduced in straightforward way from any query to the data base, (2) find out non-trivial and unsuspected relations in the data, and (3) provide insights for the development safety improvement strategies focused on PTWs. At this aim, explorative analysis of the data relative to the 254,575 crashes involving PTWs occurred in Italy in the period 2006-2008 was carried out by data mining techniques. Classification trees analysis and association rules analysis were performed. Tree-based methods are non-linear and non-parametric data mining tools for supervised classification and regression problems. They don’t require a priori probabilistic knowledge about the phenomena under studying, and no assumptions are necessary. Moreover, trees are computational feasible and consider conditional interactions among input data. Association discovery is the identification of sets of items (i.e., crash patterns) that occur together in a given event (i.e., a crash in our study) more often than they would if they were independent of each other. Thus, the association rule method can detect interdependence among crash characteristics. Results of the two analysis methods were consistent each other. Simultaneously, the two techniques present different characteristics which make their joint use complementary. Classification tree analysis provides a pictorial representation of the data and their relationships and is easily understandable. Association rule analysis provides more quantitative information, identifies the significant dependencies among all the single attributes of the data base, and allows the evaluation of the statistically power of the rules by the lift value. Analysis results showed that PTW crashes are strongly sensitive to several road, environment, and drivers attributes. The results of the analysis was successful in providing useful insight about the PTW crash patterns and their relationships. Basing on these results, engineering countermeasures and policy initiatives to reduce PTW injuries and fatalities were singled out. The use of classification trees and association rules must not, however, be seen as an attempt to supplant other techniques, but as a complementary method which can be integrated into other safety analyses.