Usability-Test der Visualisierung von Anomalien in Zeitreihen

With ongoing global digitalization, the amount of usable data for companies is increasing. Especially in the Business Intelligence field, the amount of key performance indicators that are evaluated is growing. This makes it more and more crucial to provide decision makers with the right information at the right time. Machine learning algorithms can be used to reduce the overload of information and identify abnormal tendencies. These can be used to identify anomalous occurrences of a performance indicator. Subsequently, a visualization with high usability is of high relevance to communicate the anomalies detected by the algorithm to the user in an effective, efficient and satisfying way. 

This thesis deals with the representation of anomalies in time series and investigates this exemplary on the reproduction rate R of the Covid-19 cases from the German states. Two types of visualization were analyzed: one is characterized by red markings of anomalies, another variant has a confidence interval, the range of normal data, in addition to the red markings. The variables to investigate were effectiveness, efficiency and satisfaction with the visualization method. 

For this purpose, an experiment was designed, in which the participants had to indicate the number of detected anomalies in a time series. 

The results were significantly better, and the usability of the described visualization types was increased compared to the basic visualization without markers. Both visualization types performed better in all investigated variables.

Autoren: Bohnet, Julia
Seiten: 168

Weiterführende Links:
["['', '']", "['', '']"]


Julia Bohnet

Eingetragen von

Dr. Wolfgang Gruel

Mehr zu diesem Autor