iVIS: Interpretable Interactive Visualization for User Behavior Clusters

To improve Quality of Experience (QoE) and develop new features, understanding users and making a decision to target specific user groups are important to service providers. Based on the internal interviews, we find that service operators have trouble in identifying user behavior characteristics for numerous services in a short time. To address this challenge, we present iVIS, an interactive visualization system that clusters the user behaviors and visualizes the representative behavior patterns. With iVIS, service providers can interpret the user clusters (e.g., heavy/light users), and drill down to a particular cluster to get details interactively. To evaluate our system, we conduct a case study on the log data from the internal data catalog service which enables researchers to browse and use datasets. We found service operators could rapidly interpret representative user behaviors, and discover new behavior types such as frustrated users (i.e., users who only explored datasets for a while but not use them) or testers (i.e., users who used the service for testing) by refining the clustering results.