Preview |
PDF, German
- main document
Download (2MB) | Terms of use |
Abstract
In this thesis a failure prediction environment based on selected methods of statistics, machine learning and data mining was developed. For the system behavior representation of an observed information system mathematical models,such as Bayesian networks and Markov chains, were generated. These models were analyzed using a probabilistic model-checking for both the process evolution (path) into a fault condition as well as the time duration until the occurrence of this possible fault. Furthermore, based on a Kalman filter an approach for identification of anomalies and / or misbehavior in data streams was developed. This also includes a system identification part, which derives the models from measurement data. These models form the basis for the misbehavior detection.
Document type: | Dissertation |
---|---|
Supervisor: | Reuter, Prof. Dr. Dr. h.c. Andreas |
Date of thesis defense: | 27 September 2017 |
Date Deposited: | 23 Nov 2017 10:40 |
Date: | 2017 |
Faculties / Institutes: | The Faculty of Mathematics and Computer Science > Dean's Office of The Faculty of Mathematics and Computer Science |