Publication
Leveraging Context Information in Spatio-Temporal Big Data Analytics: a Study in the Mobility Domain
Ricardo Martinez
Mastersthesis, Technische Universität Berlin, 9/2018.
Abstract
The main objective of this thesis is to evaluate the use of a Spatio-Temporal approach in order to provide contextual data in the mobility domain. The Daystream project1 aims to explore innovative techniques and tools to improve mobility experience, traffic safety, and quality of service. Daystream is funded by the Federal Ministry of Transport and Digital Infrastructure, Bundesministerium f¨ur Verkehr und digitale Infrastruktur (BMVI )2. This Master thesis is framed into the Daystream project, within a use case to improve mobility experience. Spatio-Temporal is, in addition to other features of interest, the combination of spatial and temporal dimensions, which denotes the location and time. Traffic engineering is a part of the Mobility field defining traffic parameters used in this thesis, such as Speed, Time Mean Speed, Space Mean Speed, Density, Time Headway, Distance Headway, among others. These traffic parameters are used to define Level of Service and the Traffic Congestion condition of a road. The methodology goes through a process divided into two steps: • Traffic Congestion Areas Detection: get road sensor data, calculate traffic parameters from that data, filter the concrete places of Traffic Congestions and detect areas where the density of sensors is high. • Levering Context Information: use of External Events as the context information for Traffic Congestion Areas, by calculating relationships between Traffic Congestion Areas and External Events. This thesis shows the use of a Spatio-Temporal approach in the mobility field to include External Events as the context of a Traffic Congestion Area. As the main contributions this thesis uses traffic parameters and formal definitions of Traffic Congestion, the main focus is an analysis of an area rather than a single point or road segment, the changes over time are considered, and finally the use of data not related to roads. The contextual data can be used to prediction traffic congestions impact or provide further information to take decisions on policies and strategies, and eventually to avoid or reduce such impact.