Detecting unstructured events through multi-source data analysis

There is increasing interest in using urban data analysis for different applications, particularly in relation to urban management, transport optimisation and fields related to cities becoming ‘smarter’. Data used in the analyses comes from different sources and involves various domains, from transportation engineering to urban planning. Three forms of data in particular have been of interest within the research community for a number of years. The first one includes data sourced from what are called ‘smart cards’ (for ticketing purposes): mainly used for calculating commuting flows and tracking users trajectories (Munizaga and Palma 2012; Hasan et al. 2012), it also has some applications for urban zoning and activity detection (Roth et al. 2010; Zhong et al. 2014). The second type concerns mobile phone data, generally aggregated and anonymised by providers, used mainly for detecting urban dynamics and activities because of the fine granularity, considering mobile phone as a proxy for each single user. Research focus has been similar from pioneering works till recent ones (Ratti et al. 2006; Candia et al. 2008; Calabrese et al. 2011; Hasan et al. 2012). Finally, the third one involves social media data, which use increased in latest years mainly because of the users’ growth and its availability through API. Spatial applications are particularly relevant when data attributes include location information, as seen in Twitter (Hawelka et al. 2014), or Foursquare (Cranshaw et al. 2012) data. In this paper we will discuss how a deeper understanding of urban dynamics can be yielded through the joint analysis of all three datasets. Specifically we seek to develop a platform for the identification of events, with a view to informing urban planning and management.