A Survey on Interaction Log Analysis for Evaluating Exploratory Visualizations

Abstract

The trend of exploratory visualization development has driven the visual analytics (VA) community to design special evaluation methods. The main goals of these evaluations are to understand the exploration process and improve it by recording users' interactions and thoughts. Some of the recent works have focused on performing manual evaluations of the interaction logs, however, lately some researchers have taken the step towards automating the process using interaction logs. In this paper we show the capability of how interaction log analysis can be automated by summarizing previous works' steps into building blocks. In addition, we demonstrate the use of each building block by showing their methodologies as use case scenarios, such as how to encode and segment interactions and what machine learning algorithms can automate the process. We also link the studies reviewed with sensemaking aspects and interaction taxonomies selection.