Representing unevenly-spaced time series data for visualization and interactive exploration

Abstract

Visualizing time series is useful to support discovery of relations and patterns in financial, genomic, medical and other applications. Often, measurements are equally spaced over time. We discuss the challenges of unevenly-spaced time series and present four representation methods: sampled events, aggregated sampled events, event index and interleaved event index. We developed these methods while studying eBay auction data with TimeSearcher. We describe the advantages, disadvantages, choices for algorithms and parameters, and compare the different methods for different tasks. Interaction issues such as screen resolution, response time for dynamic queries, and learnability are governed by these decisions. © IFIP International Federation for Information Processing 2005.