Data-Driven Quality of Experience Management of Prevailing Internet Applications

The Internet has become an essential part of everyday life, supporting information access, communication, news consumption, social media, and entertainment. Consequently, web browsing and video streaming dominate today's network traffic, placing significant pressure on all stakeholders involved in the service delivery chain, while end users continue to expect seamless performance and high service quality, independent of underlying technologies. Optimizing Quality of Experience (QoE) is thus essential for all stakeholders across the service delivery chain, as poor experiences might result in user churn and, ultimately, revenue loss. Thus, this thesis addresses the optimization of QoE for prevailing Internet applications in three stages. First, subjective user studies are conducted to assess the suitability of the Google Core Web Vitals for holistic web QoE models and to assess the impact of cookie consent banners on perceived quality. Second, the feasibility of data-driven QoE modeling using explainable artificial intelligence (XAI) is investigated, combining the advantages of classical modeling approaches with those of machine learning. Finally, strategies for optimizing short-form video services, such as TikTok or Instagram Reels, are explored at both the application and network levels. Overall, this work contributes to the holistic, user-centered optimization of modern Internet applications.