Visual Analytics of Streaming Data in Concept Drift Detection

Recent developments in sensor technology have enabled real-time data acquisition, high-frequency and multimodal data capturing thus underlying the need for monitoring physical or operational conditions in various aspects of the data. High dimensionality and volume of the data poses significant statistical challenges creating a need for accurate concept drift detection and prompt responses, a task that can be performed either by an algorithm or by human expert. Visual analytics plays a crucial role in concept drift detection by enabling analysts to visually explore, analyze, and interpret complex data streams, facilitating real-time monitoring, decision-making, and communication with stakeholders. This paper examines the usability of radar graphs for concept drift exploration, as well as their usage in other steps of analysis. For the purpose of demonstration, a pipeline for real-time health data analysis is presented and steps are visualized using radar graphs.