Spatial Interpolation of Traffic Data by Genetic Fuzzy System

Abstract

We propose a method to interpolate traffic data of roads using genetic fuzzy systems (GFSs). In Japan, car navigation equipment provides drivers with real-time traffic information about principal roads. The information enables giving route guidance. In a previous study, the problem of the method lies in the following two facts because a human designs membership functions of fuzzy c-means (FCM) experientially. One fact is that the design cost is high; the other is that tuning membership functions optimally is difficult. We automatically tune membership functions using a genetic algorithm (GA). The membership functions are encoded as a chromosome of GA, and the average of mean daily errors calculated from actual traffic data is used as a fitness function. Experiments using actual traffic data and an actual road map indicate that our method is more effective than the conventional method