Guided Visual Analysis of Multivariate Time Series

Abstract

Time series analysis examines the behavior of data over time. Many papers analyze single variable patterns and trends. However, few proposals have been made for analyzing various variables; this is because when the complexity of the analysis increases, the time series are not easy to understand due to their dynamic and heterogeneous nature. We can resort to Visual Analytics, which transforms complex data into simple graphs, which not only support the exploration of multivariate patterns but also assist the user in understanding found patterns. However, as the set of multivariate time series grows, the analysis methods become increasingly complex, making it difficult to obtain knowledge. In this way, we propose a new approach to visual analysis based on guides and oriented to multivariate time series, which reduces the complexity in the analysis process through guidance and direction guides. An interactive visualization tool was implemented using guides called VisWeb, which integrates the proposed guides with Visual Analysis techniques. For optimal visual analysis of multivariate time series with guides and to demonstrate the reduction of complexity, four questions were posed, designed by the domain expert, referring to the workflow of tasks, task guide, history of the sequence of steps, and limitation of visual space. The results showed that the proposed approach adequately satisfies the questions posed, allowing a reduction in complexity.