Visual Features for Multivariate Time Series
Bao D. Nguyen, Rattikorn Hewett, Tommy Dang · 2020
Visual analytics combines the capabilities of computers and humans to explore the insight of data. It provides coupling interactive visual representations with underlying analytical processes (e.g., visual feature extraction) so that users can utilize their cognitive and reasoning capabilities to perform complex tasks effectively or to make decisions. This paper applies successfulness of visual analytics to multivariate temporal data by proposing an interactive web prototype and an approach that enables users to explore data and detect visual features of interest. A list of nonparametric quantities is proposed to extract visual patterns of time series as well as to compute the similarity between them. The prototype integrates visualization and dimensional reduction techniques to support the exploration processes. Many different temporal datasets are used to justify the effectiveness of this approach, and some remarkable results are presented to show its value.