I am a PhD student at the Machine Learning Group of Prof. Dr. Ulf Brefeld at the
Leuphana University of Lüneburg.
I received a Master of Science in Mediology at the Aalborg University
and my Bachelor in Digital Media (minor: E-Business) at the Leuphana University of Lüneburg.
Research Interests
In my Ph.D. studies I am investigating mouse movement for determining users, user intention and behaviour.
The evaluation of the mouse information is usable to measure the users’ mood, draw conclusions about a users’ satisfaction, or get information about a users’ working behaviour and performance.
At this, I take a closer look on one-class approaches (esp. anomaly detection), time series analysis and cursor-motif discovery.
Understanding users and their behaviour is of high interest in Human-Computer Interaction (HCI) since this can be used to personalise the user interface and ensure a satisfying user experience.
Such personalisation increases long term user engagement and, as a side-effect, also drives up other aspects of the services, for instance, revenue.
The aim is to leverage machine learning techniques to investigate data from mouse movements to identify single users and predict the short- and long-term behaviour of users.
Teaching
Einführung in die Programmierung mit Python zur Datenanalyse (W20, W21)
Programmierung in Python (Professional School) (S21, S22)
Publications
J. J. Matthiesen, U. Brefeld. When more is less. Adverse Effects in Outlier Exposure(abstract). Proceedings of the Northern Lights Deep Learning Workshop, 2022
[Abstract]
[Poster]
International Conference on Human-Computer Interaction
Pages:
95-113
This work investigates the potential bivariate correlations between selected pattern related mouse attributes and a set of factors for the determination of the satisfaction with the usability. To examine this, a prototype tool for the analyzation and characterization of mouse attributes, Simple Mouse Attribute Analysis (SMATA), within the usage of a cloud-based vertical business software solution for managing soft data, was designed and implemented. A questionnaire was conducted to evaluate the users’ satisfaction with the usability. Following, the potential correlation between those properties was investigated. The findings revealed
...
several statistically significant correlations between the factors of satisfaction with the usability and the examined mouse attributes. Mouse attributes like the number of direct movement, the number of long direct movements, the number of made pauses, as well as the covered distance and the total time of the session could be associated with the perception of the system usefulness, the information and interface quality and the overall impression. The objective of this study was to point out a new interesting research direction of using implicit gathered user data from one of the default communication channels in HCI: the computer mouse.
show less
International Conference on Human-Computer Interaction
Pages:
68-75
In this working paper, we study user identification via mouse movement. Instead of treating the problem as a multi-class classification task, we cast user identification as a one-class problem and propose to learn an individual model for every user. Preliminary empirical results show that our approach works for some but not all users. We report on lessons learned.