This paper focuses on the discovery of unusual spatiotemporal associations across multiple phenomena from distinct application domains in a spatial neighborhood where each phenomenon is represented by anomalies from the domain. Such an approach can facilitate the discovery of interesting links between distinct domains, such as links between traffic accidents and environmental factors or road conditions, environmental impacts and human factors, disease spread, and hydrological trajectory, to name a few. This paper proposes techniques to discover spatiotemporal associations across distinct phenomena using a series of anomalous windows from each domain that represent a phenomenon. We propose a novel metric called influence score to quantify the associated influence between the phenomena. In addition, we also propose spatiotemporal confidence, support, and lift measures to quantify these associations. Two novel algorithms for finding multi-domain spatiotemporal associations across phenomena are proposed. We present experimental results across real-world phenomena that are linked and discuss the efficacy of our approach.

Researchers should cite this work as follows:

• Prathamesh Walkikar; Lei Shi; Bayu Adhi Tama; Vandana Janeja (2024), "Discovery of multi-domain spatiotemporal associations," https://theghub.org/resources/5139.

  BibTex | EndNote

1. anomalies
2. spatial neighborhood
3. spatiotemporal