Speaker
Description
Accurate modelling, prediction and representation of traffic flows is an essential element of intelligent transportation systems, urban planning, and smart environments in general. Road traffic creates a complex dynamic system with many stochastic elements and many internal and external dependencies. Road traffic monitoring can be monitored by inexpensive sensing and monitoring systems and is often readily available.
In this work, we present two methods for traffic data analysis. The first methods works with location–specific hourly traffic flows, that are represented by finite mixtures of circular normal statistical distributions. The parameters of the finite mixtures are found by differential evolution, an evolutionary algorithm that is able to fit the statistical models to data with a high level of accuracy. The second methods use a modified version of a recent machine–learning method, evolutionary fuzzy rules, to learn location–specific estimators of hourly traffic flow at specific locations.
We compare predicted models against real data. Results are depicted in special circular graph form which is aesthetically pleasing and easy-to-read for general public.
