Estimating road traffic congestion from cellular handoff information using cell-based neural networks and K-means clustering

Abstract

This research proposes alternative methods for estimating degrees of road traffic congestion by using cell dwell time (CDT) information available from cellular networks. CDT is the duration that a cellular phone remains associated to a base station between handoff events. As a phone in a vehicle travels along a road having different degrees of congestion, the value of CDT varies accordingly. Measurements of CDT were taken and classified into one of the three degrees of congestion using 1) K-means clustering algorithm and 2) backpropagation neural network. These machine-assigned classifications were then compared against human opinion to assess the accuracy. The results demonstrate the feasibility of using K-means and neural networks in classifying degrees of traffic congestion and that the neural network approach performs well for this task.