1. Introduction
- mobile phones and other ICT devices collect intra-personal behaviors (e.g., individual call frequencies) and inter-personal level behaviors (e.g., social networks).
- A typical mobile phone dataset contains three categories of user information; (i) spatiotemporal tracking information; (ii) service usage information; (iii) demographic profiles (if capable)
- However, because the data precision, accuracy, and richness vary for datasets, there have not been sufficient studies on how much information gets recorded in georeferenced mobile phone data, and to what extent researchers have utilized it.
- This paper focuses on (i) datasets instead of technologies and (ii) both "individual-oriented" and "aggregated-oriented" research.
2. Mobile Phone Tracking
- Mobile phone tracking can be divided into two techniques. (i) Lagrangian-based measurements follow a device as it moves and (ii) Eulerian-based measurements have geographically fixed observation sites.
1) GSM (Group Special Mobile)-based
- It is widely available due to the widespread and low cost of GSM facilities.
- Their location is determined by the identity (ID) of the connected base tower.
- The location precision can be improved by using timing advance (TA), received signal strength (RSS), and cell tower triangulation.
- Raw usage data include usage detail records (UDRs).
- The details of phone calls are documented as call detail records (CDR).
- However, researchers may not get complete information from CDRs because the manufacturers and service providers can decide which information is emitted and how it is formatted.
2) Wi-fi-based
- Similar to GMS tracking, this technique also tracks the network cell that a mobile device registers to.
- It can be retrieved from an Internet protocol detail record (IPDR).
- IPDRs can be encoded using extensible markup language (XML) and external data representation (XDR).
- However, they can collected by Internet service providers only.
3) Bluetooth-based
- It can be particularly useful for studying the social and spatial interactions between individuals due to their short wavelength.
- The location information can be extracted based on the relative distance between a certain device and pre-located Bluetooth receivers.
4) GPS-based
- Assisted GPS (A-GPS) or differential GPS (D-GPS) modules are usually incorporated into smartphones.
- A-GPS communicates with servers in the base tower, and the locations are calculated by servers instead of handsets.
- Inertial navigation systems in smartphones also available tracking methods.
3. A Summary of Existing Mobile Phone Datasets
1) Spatio-temporal tracking information
- The accuracy and precision of trajectories depend on the location technique, and the collected data are typically generated as sample points.
2) Service usage information
- It is information recorded in mobile phone datasets, which also includes pieces of social interaction information of mobile phone users.
3) User profiles
- The aggregated-level social-economic profiles may be obtained.
- They obtained 47 georeferenced mobile phone datasets and summarized the information following six properties: sample size, temporal duration, area covered, location technique and location accuracy, service usage type, and user and additional user attributes.
- Explanation about Table A-1:
- There are two distinct sources of data: (i) human participant experiments (ii) existing usage record data
- Study locations
- All positioning techniques have data-quality issues
- User profiles are commonly available per the agreement between researchers and participants
- Researchers have also raised ethical concerns regarding utilizing personally identifiable information (PII) in georeferenced mobile phone datasets
- Few studies have linked social interaction with spatiotemporal tracking information and investigated these two types of information in a synergistic way
4. Research Questions Based on Georeferenced Mobile Phone Data
- There are possible research questions and they can be categorized into two: individual-oriented and aggregate-oriented
- In the real world, it is also feasible to modify or extend Tables 2 and 3.
