Transportation melting pot Dhaka: road-link based traffic volume estimation from sparse CDR data

Understanding traffic conditions in urban areas is an important research direction, especially in rapidly growing cities that still struggle with congestion and inefficient traffic control strategies. The purpose of this study is to estimate the road-link scale traffic conditions of the metropolitan area of Dhaka, where a variety of transportation demands get mixed up as if it were a 'melting pot', based on sparse mobile phone data and road networks. The mobile phone data used here is the Call Detail Records (CDR). Our method extracted Origin and Destination (OD) from CDR in two ways. One is a simple extraction of continuous records with base station differences, and another is an extraction of trips between significant locations through CDR clustering and trip segmentation. Full-day link traffic volume is then estimated by assigning hourly trips with each OD to routes on actual road network. The methodology is demonstrated using 1 month CDR from 6.85 million users of Dhaka, with only 5.8 logs per day in average. Our estimation results show a relatively strong correlation (r=0.75) with the actual traffic count in a road-link scale. Moreover, the fact that the estimation results have close accuracy with the Person Trip survey data based estimation suggests that traffic conditions understanding based on long-term mobile phone data is a valid method for large-scale traffic survey.