@phdthesis{Maier2006, author = {Maier, Thilo}, title = {Modeling ETL for Web Usage Analysis and Further Improvements of the Web Usage Analysis Process}, url = {http://nbn-resolving.de/urn:nbn:de:bvb:824-opus-363}, school = {Katholische Universit{\"a}t Eichst{\"a}tt-Ingolstadt}, year = {2006}, abstract = {Currently, many organizations are trying to capitalize on the Web channel by integrating the Internet in their corporate strategies to respond to their customers' wishes and demands more precisely. New technological options boost customer relationships and improve their chances in winning over the customer. The Web channel provides for truly duplex communications between organizations and their customers and at the same time, provides the technical means to capture these communications entirely and in great detail. Web usage analysis (WUA) holds the key to evaluating the volumes of behavioral customer data collected in the Web channel and offers immense opportunities to create direct added value for customers. By employing it to tailor products and services, organizations gain an essential potential competitive edge in light of the tough competitive situation in the Web. However, WUA cannot be deployed offhand, that is, a collection of commercial and noncommercial tools, programming libraries, and proprietary extensions is required to analyze the collected data and to deploy analytical findings. This dissertation proposes the WUSAN framework for WUA, which not only addresses the drawbacks and weaknesses of state-of-theart WUA tools, but also adopts the standards and best practices proven useful for this domain, including a data warehousing approach. One process is often underestimated in this context: Getting the volumes of data into the data warehouse. Not only must the collected data be cleansed, they must also be transformed to turn them into an applicable, purposeful data basis for Web usage analysis. This process is referred to as the extract, transform, load (ETL) process for WUA. Hence, this dissertation centers on modeling the ETL process with a powerful, yet realizable model - the logical object-oriented relational data storage model, referred to as the LOORDSM. This is introduced as a clearly structured mathematical data and transformation model conforming to the Common Warehouse Meta-Model. It provides consistent, uniform ETL modeling, which eventually supports the automation of the analytical activities. The LOORDSM significantly simplifies the overall WUA process by easing modeling ETL, a sub-process of the WUA process and an indispensable instrument for deploying electronic customer relationship management (ECRM) activities in the Web channel. Moreover, the LOORDSM fosters the creation of an automated closed loop in concrete applications such as a recommendation engine. Within the scope of this dissertation, the LOORDSM has been implemented and made operational for practical and research projects. Therefore, WUSAN, the container for the LOORDSM, and the LOORDSM itself can be regarded as enablers for future research activities in WUA, electronic commerce, and ECRM, as they significantly lower the preprocessing hurdle - a necessary step prior to any analysis activities - as yet an impediment to further research interactions.}, subject = {Data Mining}, language = {en} }