Time series form an important data source for Business Intelligence (BI), and are collected in a large variety of application domains e.g. sales, energy consumption or manufacturing. Due to the ongoing trend of data gathering, time series not only increase in length but also in numbers and variety, making analysis techniques like data clustering a necessity. From an end user’s perspective, this brings major difficulties. First, time series are a challenging format as they do not represent static discreet objects but a dynamic development over time. Second, the wide variety of available clustering algorithms and their technical parameterization make clustering very user-unfriendly. This topic aims to overcome this issue by developing a feedback-driven analysis process. Based on the concepts of the Research Project AUGUR, two process components should be designed. First, a backend algorithm management that covers data handling, preprocessing and algorithm execution for time series analysis. Second, a visual-interactive frontend, that displays results, guides the user, and offers intuitive and user-friendly feedback operations.

Main Advisor at Technische Universität Dresden (TUD)
Co‐advisor at Aalborg Universitet (AAU)