Mining fuzzy rules for a traffic information system

Abstract

This work presents a fuzzy system for pattern recognition in a real application: the selection of traffic information messages to be displayed in Variable Message Signs located at the main routes of the city of Rio de Janeiro. In this application, flow and occupancy rate data is used to fit human operators' evaluation of traffic condition, which is currently done from images of strategically located cameras. The fuzzy rule-base mining is presented considering the symbolic relationships between linguistic terms describing variables and classes. The application presents three classifiers built from data.