Leveraging spatio-temporal clustering for participatory urban infrastructure monitoring

Internet-enabled, location aware smart phones with sensor inputs have led to novel applications exploiting unprecedented high levels of citizen participation in dense metropolitan areas. Especially the possibility to make oneself heard on issues, such as broken traffic lights, potholes or garbage, has led to a high degree of participation in Urban Infrastructure Monitoring. However, duplicate reporting by citizens leads to bottlenecks in manual processing by municipal authorities. Spatio-temporal clustering can serve as an essential tool to group and rank similar reports. Current data mining techniques could be used by municipal departments for this task, but the mandatory parameter selection can be unintuitive, time consuming and error-prone. In this work, we therefore present a novel framework for clustering spatio-temporal data. We first apply an intuitive transformation of the data into a graph structure and subsequently use well-established parameter-free graph clustering techniques to detect and group spatio-temporally close reports. We evaluate our method on two real-world data-sets from different mobile issue tracking platforms. As one of the datasets includes labels for duplicate reports, we can show how our framework outperforms existing techniques in our exemplary use-case (duplicate detection).