On Tuesday, the 24th of July, Sarah Freytag defended her master thesis on the topic “Keystroke Analysis for Stress Detection in Business Communication and Documentation”. The chair of business information systems congratulates the student on her master’s degree.
Abstract:
Currently, the University of Rostock is working together with social support organisations on a project for better digital support for the employees, but with the increasing work- load the management suspects correspondingly increased stress. This thesis evaluates the possibility and effectiveness of non-intrusive stress detection through keystroke analysis in an organisational context. Furthermore, the clustering of data along stable personality traits for classification enhancement is investigated. For this purpose, the Dwell Time, Latency and Flight of 10 Di- and Trigraphs each were collected from employees of an organisation as well as from volunteers. Unfortunately, the data did not contain enough "stressed"data sets to train any classifiers. Instead, the reported states of valence, arousal and dominance are classified using k-Nearest Neighbour, Bayesian Network and Support Vector Machine algorithms implemented in the Waikato Environment for Knowledge Analysis (Weka). The results from the volunteer data show that arousal is difficult to classify, while valence and arousal achieve better accuracies. The data collected from employees, on the other hand, excels in valence and arousal classification with up to 79% accuracy. Grouping the data was not possible on the employee data record, but it improved the classification results for the volunteers depending on the group and classifier. Further studies with more extensive data records need to be conducted to validate these findings.