To investigate the relation between people's travel behavior and urban traffic congestion

and provide technical support to alleviate urban traffic congestion

a data-driven source localization approach was constructed based on high-coverage-rate low-precision mobile phone data and high-precision taxi GPS data. Travel demand and traffic condition information was obtained by personal mobile phone data and taxi GPS data

respectively. Mobile phone data and taxi GPS data were combined to estimate origin-destination matrix

conduct traffic assignment

and dynamically locate traffic sources of road segments and the congestion sources of the city. Result showed that a majority of traffic flow was generated by a few sources

and the sources became more concentrated when traffic jam occurred. Urban congested sources were affected by commuting behaviors

and exhibited different characteristics during morning and evening peak hours. Traffic source location by data fusion can uncover the internal mechanism and evolution law of traffic congestion

and help in making target policies dealing with traffic congestion.