Vehicle Accident Severity Rule Mining Using Fuzzy Granular Decision Tree

Abstract

Road collisions are disasters that constitute a major cause of disability and untimely death. Therefore, the need for investigation of the conditions of road collisions and driver awareness on highways is critical. A great deal of huge data, with regards to road collisions such as collision properties, road conditions, temporal information, environmental attributes, spatial measures and road geometry have been accumulated. This thesis proposes a new fuzzy granular decision tree to generate road collision rules to apply to the discrete and continuous data stored in collision databases. To improve the efficiency of the algorithm, the fuzzy rough set feature selection is applied .The major highways in California are considered as a case study to examine the proposed approach. The experimental results demonstrate that the proposed method is more accurate and efficient than the traditional decision tree methods, with the less redundancy in constructing the decision tree.