There has been growing interest in exploiting cellular network data for transportation planning purposes in recent years. In this paper, we utilize these data for determining mode of travel in the city of Shiraz, Iran. Cellular data records -including location updates in 5minute time intervals- of 300,000 users from the city of Shiraz has been collected for 40 hours in three consecutive days in a cooperation with the major telecommunications service provider of the country. Depending on the density of mobile BTS’s in different zones of the city, the user location can be located within an average of 200 meters. Considering data filtering and smoothing, data preparation and converting them to comprehensible traces is a large portion of the work. A novel approach to identify stay locations is proposed and implemented in this paper. Origin-Destination matrices are then created based on trips detected, which shows acceptable consistency with current O-D matrices. Finally, Travel times for all trips of a user is estimated as the main attribute for clustering. Trips between same origin and destination zones are combined together in a group. Using K-means algorithm, records within each group are the portioned in two or three clusters, based on their travel speeds. Each cluster represents a certain mode of travel; walking, public transportation or driving a private car.
More detail can easily be written here using Markdown and $\rm \LaTeX$ math code.