Abstract
The understanding of visual analytics process can benefit visualizationresearchers from multiple aspects, including improving visual designs anddeveloping advanced interaction functions. However, the log files of userbehaviors are still hard to analyze due to the complexity of sensemaking andour lack of knowledge on the related user behaviors. This work presents a studyon a comprehensive data collection of user behaviors, and our analysis approachwith time-series classification methods. We have chosen a classicalvisualization application, Covid-19 data analysis, with common analysis taskscovering geo-spatial, time-series and multi-attributes. Our user study collectsuser behaviors on a diverse set of visualization tasks with two comparablesystems, desktop and immersive visualizations. We summarize the classificationresults with three time-series machine learning algorithms at two scales, andexplore the influences of behavior features. Our results reveal that userbehaviors can be distinguished during the process of visual analytics and thereis a potentially strong association between the physical behaviors of users andthe visualization tasks they perform. We also demonstrate the usage of ourmodels by interpreting open sessions of visual analytics, which provides anautomatic way to study sensemaking without tedious manual annotations.