Categorizing Uncertainties in the Process of Segmenting and Labeling Time Series Data

Abstract

The segmenting and labeling of multivariate time series data is applied in different domains, e.g. activity recognition or sensor states. This involves several steps of (pre-) processing, segmenting, and labeling of time intervals, and visually exploring the results as well as iteratively refining the parameters for all the processing steps. Within these processes different uncertainties are involved and relevant. In this poster we identify and categorize important uncertainties in this problem domain. We discuss challenges for visually communicating these uncertainties throughout the segmenting and labeling process.