Big Data Research, Volume 15 (pp. 12-28)
Abstract. Geolocated time series are time series that correspond to specific locations. They can represent, for example, visitor check-ins at certain venues or readings of sensors installed at various places. The amount and significance of such time series have increased in many domains over the last years. However, although several works exist for time series visualization and visual analytics in general, there is a lack of efficient techniques for visual exploration and analysis of geolocated time series in particular. In this paper, we present two approaches that rely on hybrid spatial-time series indices to allow for efficient map-based visual exploration and summarization of geolocated time series data. In particular, we use the BTSR-tree index and we introduce a new variant of the iSAX index, called geo-iSAX. The former is a spatial-first hybrid index that extends the R-tree by maintaining bounds for the time series indexed at each node. Following a similar rationale, geo-iSAX is a time series-first hybrid index that maintains spatial MBRs of the geolocated time series indexed in each node. We describe the structure of these indices and show how they can be directly exploited to produce map-based visualizations of geolocated time series at different levels of granularity. We empirically validate our approach using two real-world datasets, as well as a synthetic one that is used to test the scalability of our methods.