Classifying events in transportation and road safety. Some algorithms and models
ARON ; SEIDOWSKY
Type de document
COMMUNICATION AVEC ACTES INTERNATIONAL (ACTI)
Langue
anglais
Auteur
ARON ; SEIDOWSKY
Résumé / Abstract
This paper describes a help to the diagnosis, which consists in classifying events (as road accidents) in pre-defined classes. Different sources of knowledge of the area, statistics, logical rules, questions to the operator. Uncertainty and inaccuracy are processed to some extend. The tool 'crique' (classifying from rules, informatics data, questions), which has been used for several applications, is presented. The interest of different algorithms coming from our team or from others is highlighted. A method for the creation of the classes is described. An algorithm for the choice of the next question is reported. An algorithm of reconstruction of missing data is used to propose default values for the user.
