This master thesis consists of three separate articles, all focused on the hybrid approach to web usage mining that combines an aggregated structure and a data warehouse schema to enable flexible, constraint-based extraction of knowledge. The first article presents a case study in which a framework for extracting knowledge from the use of an interactive story is developed. The paper describes the hybrid approach, presents an overview of the developed framework called MIMER and illustrate principle user interfaces to the framework. The second paper presents an expansion of the hybrid approach called PEHA in which the knowledge extracted is validated against the entire usage data stored in a data warehouse. Experiments show that the knowledge can be validated in near constant time and is competitive compared to a rival data warehouse schema when considering running time, flexibility and storage requirements. The third paper investigates the quality of the knowledge extracted from the aggregated structure, considering the similarity and accuracy of the extracted knowledge. Experiments indicate that the extracted knowledge is relatively high but drops as the size of the knowledge grow compared to the precision of the aggregated structure.