Structure-Based Rule Selection Framework for Association Rule Mining of Traffic Accident Data

Abstract

A rule selection framework is proposed which classifies, selects, and filters out association rules based on the analysis of the rule structures. It was applied to real traffic accident data collected from local police stations. The rudimentary nature of the data required several passes of association rule mining to be performed, each with different sets of parameters, so that semantically interesting rules can be spotted from the pool of results. It was shown that the proposed framework could find candidate rules that offer some insight into the phenomena being studied.