Author(s): Vasileios Asthenopoulos, Ioannis Loumiotis, Pavlos Kosmides, Evgenia Adamopoulou, Konstantinos Demestichas
Abstract: This paper presents the concept of using mobile network performance data in order to estimate and predict road traffic conditions. The effectiveness of the approach taken by the authors is examined using realworld data acquired from mobile and road network operators. Furthermore, a comparative analysis is performed to evaluate which of the two machine learning techniques proposed, namely the Multi-Layer Perceptron and the General Regression Neural Network is more suitable for this purpose. It is argued that practical implementations of the system described in this paper can reduce the number of sensors needed to acquire metrics from the road network, allow accurate estimation of future road traffic conditions exclusively using anonymous mobile network performance data, and even raise near real-time alerts about traffic events, without the requirement of dedicated traffic sensors.