Using Topological Analysis to Support Event-Guided Exploration in Urban Data

Fig. 1. Overview of the event guided exploration technique. First, (1) the input data is transformed into a time-varying scalar function. (2) Topological features are computed from the scalar functions to identify the set of events. (3) An event group index is then created from the identified events to support efficient querying over a large number of events. (4) A visual interface guides the user towards interesting events (5) in the data, allowing them to select an event and (6) interactively search for similar events. Abstract — The explosion in the volume of data about urban environments has opened up opportunities to inform both policy and administration and thereby help governments improve the lives of their citizens, increase the efficiency of public services, and reduce the environmental harms of development. However, cities are complex systems and exploring the data they generate is challenging. The interaction between the various components in a city creates complex dynamics where interesting facts occur at multiple scales, requiring users to inspect a large number of data slices over time and space. Manual exploration of these slices is ineffective, time consuming, and in many cases impractical. In this paper, we propose a technique that supports event-guided exploration of large, spatio-temporal urban data. We model the data as time-varying scalar functions and use computational topology to automatically identify events in different data slices. To handle a potentially large number of events, we develop an algorithm to group and index them, thus allowing users to interactively explore and query event patterns on the fly. A visual exploration interface helps guide users towards data slices that display interesting events and trends. We demonstrate the effectiveness of our technique on two different data sets from New York City (NYC): data about taxi trips and subway service. We also report on the feedback we received from analysts at different NYC agencies. Index Terms—Computational topology, event detection, spatio-temporal index, urban data, visual exploration.