01660390

Component

The use of social network big data sources has attracted significant research attention to understand urban mobility patterns. In this paper, we explore the potentials of using the Location-based Social Network (LBSN) for Dynamic Urban Origin-Destination (OD) estimation. The existing dynamic OD models can be categorized based on the data resource include flow calibration data, vehicle re-identification data, GPS survey data, and cellphone location based data. These survey data may put an intensive load on surveyees and equipment to maintain large sample size, sample resolution, and spatial-temporal coverage. Location-based Social Networking (LBSN) data are generated when users check-in to a POI (Point of Interests), such as a restaurant, a workplace, and even a transportation facility, for social sharing and receiving venue promotions. Based on previous studies, LBSN data is found to be a useful secondary data source to estimate static and time-of-day urban travel demand. This paper uses the Pearson product-moment correlation coefficients to compare the arrival patterns of zonal LBSN check-in observations between origins and destinations. Such correlation coefficients among different OD pairs are then integrated into the friction function in a modified temporal gravity model for dynamic trip distribution process. The proposed model is applied to check-in data collected from Manhattan Island, New York City archived by Foursquare and calibrated with agency OD data and TOD factors from the New York Metropolitan Transportation Council (NYMTC). The predicted time-of-day OD flow patterns achieved mean absolute error value of 5.2, mean absolute percentage error value of -2.64%, R2 value of 0.78 in the regression analysis for OD matrices comparison, and coincidence ratio value of 0.86 for trip length distribution comparison. Furthermore, the land use based time-of-day OD flow patterns between residential area, commercial area, transportation hub area, and open space area were reported. The results show the potentials of applying the spatial-temporal correlation analysis method for dynamic urban travel demand estimation on anonymized data.

This paper was sponsored by TRB committee ADB20 Standing Committee on Effects of Information and Communication Technologies (ICT) on Travel Choices.

18-03032

English

19p

2018

Transportation Research Board 97th Annual Meeting

Digital/other

Figures; Maps; References; Tables

Data analysis; Origin and destination; Social media; Stopped time delays

Manhattan

Data and Information Technology; Highways; Operations and Traffic Management; Planning and Forecasting

Transportation Research Board Annual Meeting 2018 Paper #18-03032

TRIS, TRB, ATRI

Jan 8 2018 10:43AM