Jung, ReinerJürgensen, LarsKelter, Udo2022-11-242022-11-242020https://dl.gi.de/handle/20.500.12116/39783Engineers use workload models to estimate the future utilization of services, understand usage profiles of services, and drive plans to evolve software systems. We usually separate workloads in intensities and user behavior models in context of performance evaluations and forecasting. Outside of software engineering user behaviors are often collected in clickstreams. They may represent the complete history of a user at one site. Ideally, a number of clickstreams can be transformed into a user behavior model containing multiple typical behavior pattern usable for a workload model. While the collection of user behaviors is fairly simple, the extraction of behavioral patterns is complicated. Current approaches are tailored for specific domains, are only applicable for one purpose, and refuse to be understood and analyzed in a meaningful way. In this paper, we introduce our latest advances to identify a suitable identification approach and layout further obstacles which must overcome to have comprehensible user behavior models.enserviceperformanceuser behaviorforecastbehavior patternText/Conference Paper0720-8928