Detecting and imaging irregularities in time-series data
Qian Zhang, Feng Lin, Hock Soon Seah · 2018 International Workshop on Advanced Image Technology (IWAIT) · 2018
Imaging and visual analytics are of great importance for problems that need closely coupled human and machine analysis. In this paper, we propose an interactive system to show irregularities in a time-series dataset. The key technique is a bar-chart-like irregularity plot that gives user a quick insight of the entire time series dataset, with detected status such as normal, missing value, extreme value and possible outlier marked in different colors. The timestamp alignment plot that shows time-related changes and trending information can be used to evaluate patterns and validate automatic detection result in irregularity plot. Technical descriptions of the detection methods and results are presented. Data analysts can benefit from the dataset overview provided by the system before proceeding further data cleansing operations.