Cell Phone Data and Travel Behavior Research: Symposium Summary Report

The Exploratory Advanced Research Program

Agency and Academic Experience Using Cell Data

In this next session, researchers from various agencies and academic institutions provided descriptions of the work they have been conducting and their uses of cellular data. Symposium participants then discussed challenges relating to sampling and dataset extrapolation.

Presentations

Cynthia Chen, University of Washington
Cynthia Chen identified certain challenges and limitations associated with the use of cell data. These include difficulties with penetration rates and sample sizes (e.g., in some markets users have multiple phones), as well as the varying market shares of different carriers. Another challenge is oscillation between two cell towers and therefore being unable to determine an exact location. In a preliminary dataset of 8,000 people, Chen was able to determine 90 percent of homes and workplaces within a certain area. Current key challenges that she and her team are working on include validation of trips and population expansion.

Xiaowei Xu, University of Arkansas
Xiaowei Xu conducted research that focuses on determining data clusters and associating individuals with groups of people or communities. He accomplished this through an algorithm that connects individuals’ social networking activities, as shown in figure 3. One specific example where Xu applied this algorithm was at college football games. Xu noted that challenges arise when cleaning and using data, as well as when finding a dynamic algorithm to enhance the result. Xu told the symposium participants that a major area to improve upon is to establish new methods for validating travel behavior by using cell data.

Figure 3. Clustering algorithm developed to represent individuals in a complex system.

Elizabeth Birriel, Florida Department of Transportation
Elizabeth Birriel explained that the Florida Department of Transportation (FDOT) has a desire to use cellular data to improve information used in 511 systems. These systems provide information about travel systems and travel routes. This information improvement was already achieved by using 2010 data, but it would be helpful to continue to refine these systems through the use of more data as coverage was strong for the interstate highway system but spotty for arterial roadways. FDOT is also looking at how cellular data and automated vehicles can be leveraged with each other moving forward.

Camille Kamga and Anil Yazici, City University of New York
Camille Kamga and Anil Yazici noted that they have used cellular data for mobility studies, specifically within New York City. One study involved using taxi data and determining that there is higher demand for taxis during inclement weather, as shown in figure 4. Data were also used to determine, as others have found, that the morning and afternoon peak periods are not symmetric. Further challenges and areas of research include determining the reason for these findings, for example, if changes in mobility are based on trip cancellations or mode shifts.

Figure 4. Taxi data used to determine demand for taxis during inclement weather.

Andrew Rohne, Ohio–Kentucky–Indiana Regional Council of Governments
Andrew Rohne used data from AirSage to attempt to model trips in his region of interest (greater Cincinnati, OH). In doing so, some challenges were identified regarding counts and capacities. Rohne compared his results using AirSage data to Ohio Department of Transportation and NHTS data to see how results matched and what the discrepancies were. Lessons learned and areas of future research include thinking through which data are available and what time of year the data would be most desirable for research purposes. Rohen also noted the possibility of incorporating cellular data in an NHTS expansion.

Krishnan Viswanathan, CDM Smith
Krishnan Viswanathan explained that he used data from AirSage to attempt to model trips in his region of interest (Wilmington, NC), as shown in figure 5. He derived various trips and examined different studies on non-home, residential, and home-to-home–based trips. To better understand movement and trips, Viswanathan noted that a possible next step would be to recruit users to willingly provide location data and demographic information.

Figure 5. AirSage cellphone data sample.

Major Themes Discussed

The symposium participants discussed several major themes during this session, as follows:

• Determining a timeframe. It is important while conducting research of this nature to determine what periods of the year will be most useful to analyze data, as data are typically purchased in monthly blocks.
• Focusing on long-distance travel. Although it is possible to determine long-distance travel to some degree, it is generally not the focus of regional level analysis as the possible long-distance traveler simply moves though the region without stopping.
• Accessing different data sets. Data consists of the movement of people rather than trips in the traditional sense. It is possible to access data that are extrapolated at the census tract level or to have unextrapolated data sets, such as device movements and a mobile survey.

Key Takeaways

The symposium participants noted several key takeaways during the discussion, as follows:

• Validating results is key to determining the usefulness and meaningfulness of data.
• An important challenge moving forward is linking data to behavior through demographic information.
• Although it is possible to determine long-distance travel and mode based on duration, location, and land use, it remains inexact.