Dornieden, Christoph (2017) Knowledge-Driven User Behavior Model Extraction for iObserve. (Master thesis), Kiel University, Kiel, 85 pp.

Preview

Text thesis.pdf - Published Version Download (1MB) | Preview

Abstract

Modern cloud-based software systems are exposed to constant alterations due to changing requirements. These changes can be based on various reasons. Ever-changing usage patterns of the systems user-base can be one reason for changes. A shift in the user-behavior can result in increasing load on single services or have strong economic effects for e-commerce platforms. If the usage patterns are known, the software system can be analyzed and adapted.

In this thesis we provide an approach to extract user-behavior from monitored operation-calls of a software system. Since user-behavior is not only dependent on the navigational pattern of the user in the system, but also on the specific information processed by the call, our focus lies on adding this information to the monitoring records. We then propose an identification of user-behavior models and clustering process for similar user-behavior patterns. These user-behavior models are using the enriched call information to improve accuracy.

We implemented our approach as a pipe-and-filter based service, which is integrated into the iObserve framework. The framework provides access to the monitoring records of the system and transforms them into user-sessions. To get the call-information from the system, we extend the monitoring records to hold this data. In our service, we prepare the sessions containing user-behavior and call-information for the clustering. Then the sessions are aggregated to behavior models by a clustering algorithm.

Actions (login required)

View Item