[en] The importance of data in transportation research has been widely recognized since it plays a crucial role in understanding and analyzing the movement of people, identifying inefficiencies in transportation systems, and developing strategies to improve mobility services. This use of data, known as mobility analysis, involves collecting and analyzing data on transport infrastructure and services, traffic flows, demand, and travel behavior. However, traditional data sources have limitations.
The widespread use of mobile devices, such as smartphones, has enabled the use of Information and Communications Technology (ICT) to improve data sources for mobility analysis. Mobile crowdsensing (MCS) is a paradigm that uses data from smart devices to provide researchers with more detailed and real-time insights into mobility patterns and behaviors. However, this new data also poses challenges, such as the need to fuse it with other types of information to obtain mobility insights.
In this thesis, the primary source of data that is being examined and leveraged is the popularity index of local businesses and points of interest from Google Popular Times (GPT) data. This data has significant potential for mobility analysis as it overcomes limitations of traditional mobility data, such as data availability and lack of reflection of demand for secondary activities.
The main objective of this thesis is to investigate how crowdsourced data can contribute to reduce the limitations of traditional mobility datasets. This is achieved by developing new tools and methodologies to utilize crowdsourced data in mobility analysis.
The thesis first examines the potential of GPT as a source to provide information on the attractiveness of secondary activities. A data-driven approach is used to identify features that impact the popularity of local businesses and classify their attractiveness based on these features. Secondly, the thesis evaluates the possible use of GPT as a source to estimate mobility patterns. A tool is created to use the crowdness of a station to estimate transit demand information and map the precise volume and temporal dynamics of entrances and exits at the station level. Thirdly, the thesis investigates the possibility of leveraging the popularity of activities around stations to estimate flows in and out of stations. A method is proposed to profile stations based on the dynamic information of activities in catchment areas. Through this data, machine learning techniques are used to estimate transit flows at the station level. Finally, this study concludes by exploring the possibility of exploiting crowdsourced data not only for extracting mobility insights under normal conditions but also to extract mobility trends during anomalous events. To this end, we focused on analyzing the recovery of mobility during the first outbreak of COVID-19 for different cities in Europe.