An Interactive Visualization System for Streaming Data Online Exploration

Fengzhou Liang, Fang Liu, Tongqing Zhou, Yunhai Wang, Li Chen

Abstract: The practices of understanding real-world data, in particular the high dynamic streaming data (e.g., social events, COVID tracking), generally relies on both human and machine intelligence. The use of mobile computing and edge computing brings a lot of data. However, we identify that existing data structures of visualization systems (a.k.a., data cubes) are designed for quasi-static scenarios, thus will experience huge efficiency degradation when dealing with the ever-growing streaming data. In this work, we propose the design and implementation of an enhanced interactive visualization system (i.e., Linkube) based on novel structure and algorithms support, for efficiently and intelligibly data exploration. Basically, Linkube is designed as a multi-dimensional and multi-level tree with spatiotemporal correlated knowledge units linked into a chain. Interested knowledge aggregations are thus attained via efficient and flexible sequential access, instead of dummy depth-first searching. Meanwhile, Linkube also involves a smart caching mechanism that adaptively reserves some beneficial aggregations. We implement Linkube as a web service and evaluate its performance with four real-world datasets. The results demonstrate the superiority of Linkube on response time (\(\sim \)25% \(\downarrow \)) and structure updating time (\(\sim \)45% \(\downarrow \)), compared with state-of-the-art designs.