Transportation mode inference from anonymized and aggregated mobile phone call detail records

Abstract

Abstract-Transportation mode inference is an important research direction and has many applications. Existing methods are usually based on fine-grained sampling --collecting position data from mobile devices at high frequency. These methods can achieve high accuracy, but also incur cost and complexity in terms of the system implementation and computational resource requirements. Finally, fine-grained sampling is not always available, especially for large-scale deployment. This paper proposes a novel method to infer transportation mode based on coarse-grained call detail records. The method allows estimating the transportation mode share from a given origin to a given destination, looking also at how the share changes over time. The method can achieve acceptable accuracy with trivial cost and complexity and is suitable for the statistical analysis on transportation modes of a large population. The method can also be used as a complementary tool in situations where fine-grained sampling is unavailable or the balance between accuracy and complexity is critical. A case study using real call detail records data for the city of Boston shows the performance of the proposed method. I. INTRODUCTIONode of transportation specifies one of different kinds of transport facilities that are used to transport people, such as cars, buses, bicycles, and even walking. Transportation mode inference is a tool to determine the transportation mode of an individual traveler or a group of travelers, based on the speed, travel time or other information that can be collected from their trips. This tool has been used to provide traveling services, manage transportation and plan cities.The research on transportation mode inference has a history of more than a decade. At early stage, the technology was studied in the field of pervasive or ubiquitous computing, where the computation needs to understand the context. The context includes human activities, such as walking or driving, when the computation is being done. Body worn sensors (sensors placed in one or more positions on the body) are the major data source; see for instance [1]. However, since body worn sensors are not widely available, many research efforts tried to adopt mobile sensors, such as mobile phone or GPS, as data collection devices. Therefore, existing methods for transportation mode inference can be divided into two categories according to the data collection devices. A. Mobile phone based methodsMobile phone based methods infer the transportation mode by estimating the mobile phone's speed. The speed can be estimated by measuring the low level signals from the GSM network, such as the variance of GSM signal strength or the switch rate of cells. Generally, the variance of signal strength and the number of connected cells are greater when the mobile phone is moving faster. When the speed is within a range, it is believed that the mobile phone user is in a specific transportation mode. Since these methods are based on the speed, they can hardly distinguish transportation mode...

Discovering urban activity patterns in cell phone data

“…A stay is detected if the position estimates stay within the given radius for at least the predetermined time. A similar method to extract trips from cell phone records was described in Wang et al (2010). Trajectories of moving objects can be filtered and interpolated by assuming constraints on velocity or acceleration.…”

“…A simple method to extract trips from cell phone records was described in Wang et al (2010) where consecutive location measurements are clustered according to their geographical distance and the resulting clusters are then used as origins and destinations of trips. While the clustering method accounts for moderate noise in the cell phone track, it does not include any trajectory filtering to cope with outliers resulting from occasional large positioning errors.…”

Travel Mode Detection Exploiting Cellular Network Data

“…While fine-grained sampling may not always be available, in [8], the authors used coarse-grained mobile phone call detailed records to infer transportation mode. Travel times between pairs of defined origins and destinations were estimated -for weekdays and weekends separately-and travelers were clustered into three subgroups using k-means algorithm: walking, public transit and driving cars.…”

“…A second approach to solving the problem is to neglect the effect of time of the day on traffic speed and combine all the records of trips between origin O and destination D for all 24 hours together. Records from different weekdays can also be added as done in [8]. The problem with this approach is the fact that travel speed during different hours of the day varies significantly, especially in central congested zones.…”

Caracterización De La Movilidad Metropolitana De Málaga Mediante La Telefonía Móvil. Estudio Comparativo Frente a Encuestas

“…Un análisis exhaustivo de la información capturada por la telefonía móvil podría permitir estimar parcialmente estas características bajo ciertas hipótesis. De hecho, recientemente se han publicado trabajos en la literatura capaces de estimar detalles como el modo de transporte (Wang et al, 2010) o viajes O-D por motivo (Alexander et al, 2015) a partir de datos móviles; sin embargo, las técnicas experimentadas no se encuentran suficientemente desarrolladas para garantizar una explotación de amplio espectro. En este estudio, sin embargo, la matriz derivada de la telefonía móvil usada en el proyecto ha sido un resultado final ya procesado, garantizando la total privacidad de los usuarios de telefonía, al ser el propio operador el encargado de todo…”