An introduction to spatiotemporal data

In Chapter 2, we discussed some aspects of spatial and temporal data that make such data difficult to wrangle, analyze, and scale with statistical and machine learning methods. Recall that both types of data are not independent, with dependencies crossing either space or time in the dataset. Alternative approaches to prediction are needed, and representations of spatial and temporal data as a network allow analysts to leverage the tools of network science to find change points, mine for patterns, and even predict future system behavior in the case of temporal data. In the previous chapters, we’ve learned how to analyze both spatial and temporal data with network science tools such as centrality to understand important trends.

Many analytic tasks in the real world involve spatiotemporal data—including the analysis of world stock markets influenced by local policies, data mining of store data across locations for trends in customer...