The development of on-board car safety systems requires an accidentology knowledge base for the development of new functionalities as well as their improvement and evaluation. The Knowledge Discovery in accident Database (KDD) is one of the approaches allowing the construction of this knowledge base. However, considering the complexity of the accident data and the variety of their sources (biomechanics, psychology, mechanics, ergonomics, etc.), the analytical methods of the KDD (clustering, classification, association rules etc.) should be combined with expert approaches. Indeed, there is background knowledge in accidentology which exists in the minds of accidentologist experts and which is not formalized in the accident database. The aim of this paper is to develop a Knowledge Representation Model (KRM) intended to incorporate this knowledge in the KDD process. The KRM is implemented in a knowledge-based system, which provides an expert classification of the attributes characterizing an accident. This expert classification provides an efficient tool for data preparation in a KDD process. Our method consists of combining the modeling systemic approach of complex systems and the modeling cognitive approach KOD (Knowledge Oriented Design) in knowledge engineering.

This content is only available via PDF.
You do not currently have access to this content.