Geoscience Reference
In-Depth Information
play an enhanced role in supporting “clique-like” and modular groups instead of
being a mixing pot for many groups. We expect the downtown area to host tight-
knit social groups who do not venture to the suburbs often. (4) Finally, we expect
that suburban POIs will accommodate individuals from diverse social groups, as
these agents are likely visit different parts of the city using automobiles.
This chapter proceeds as follows. We first describe the study area and the setting
of the CDR dataset. We then describe, in the methods section, how we delineate
each user's activity spaces. We analyze how
linked activity spaces
(LAS) are
spatially correlated in an urban environment by shared points of interest (POIs). We
conclude with a discussion of the usefulness of this method, its drawbacks, potential
applications, and future work.
13.2
Study Area and Dataset
Our study area is the city of Jiamusi, located in northeastern China, with a
population 2.5 million (est. 2010). This industrial city serves as a producer of
wood pulp and newsprint and participates in the global economy via a thriving
international trade harbor. The urban core of Jiamusi is nearly 18 by 10 km in spatial
extent, and its residents travel on average 1 km a day (Kang et al.
2012
).
13.2.1
Dataset and Sampling
We focus on calls made within the city area and exclude long-distance calls. We
use a CDR (call data record) dataset of mobile cell phone calls from an undisclosed
mobile phone provider in China.
The original CDR dataset contains nearly 424,000 users over 31 days. Users are
anonymized in the dataset. Combined, users make an average of 1,600,000 calls
daily. In the 31-day time span of our dataset, each user participates in an average
of 328 calls for a total duration of 6.15 hours. Each record of a mobile phone call
contains the start time, call duration, and locations of the caller and receiver. The
locations are geo-referenced to one of 96 cell towers closest to the mobile phone's
location (Table
13.1
). The dataset does not include text messages (e.g., SMS).
We process the dataset into two parts: a social network of agents (social network
in Table
13.1
) and the activity spaces of each agent (spatial patterns in Table
13.1
).
We filter the network by including only those who use at least three cell towers
during the study period in order to eliminate users who may be confined to their
home and thus interact with the city differently than a typical mobile user. Also,
an individual may have multiple mobile phones, and a phone with fewer than
three cell towers used may represent a “secondary” or less frequently used device.
In the social network, the
number of calls
is determined between a unique pair of
users, and
duration
is the sum of call time between the two users. The network
Search WWH ::
Custom Search