Pervasive Computing is a new area of research with increasing prominence; it is situated at the intersection between human/computer interaction, embedded and distributed systems and networking technology. Its declared aim is a holistic design of computer systems, which is often described as the disappearance of computer technology into the periphery of daily life. One central aspect of this vision is a partial replacement of explicit, obtrusive interfaces for human/computer interaction that demand exclusive user attention with implicit ones embedded into real-world artifacts that allow intuitive and unobtrusive use. This kind of interaction with computer systems suits human users better, but necessitates an adaption of such systems to the respective context in which they are used. Context is, in this regard, understood as any information about the current situation of a person, place or object that is relevant to the user interaction. Context-based interaction, which is pursued by the design and implementation of context-sensitive systems, is therefore one of the building blocks of Pervasive Computing. Within the last five years, a number of seminal publications on the recognition of current context from a combination of different sensors have been written within this field. This dissertation tackles the next logical step after the recognition of the current context: the prediction of future contexts. The general concept is the prediction of abstract contexts to allow computer systems to proactively prepare for future situations. This kind of high-level context prediction allows an integral consideration of all ascertainable aspects of context, in contrast to the autonomous prediction of individual aspects like the geographical position of the user. It allows to consider patterns and interrelations in the user behavior which are not apparent at the lower levels of raw sensor data. The present thesis analyzes prerequisites for user-centered prediction of context and presents an architecture for autonomous, background context recognition and prediction, building upon established methods for data based prediction like the various instances of Markov models. Especial attention is turned to implicit user interaction to prevent disruptions of users during their normal tasks and to continuous adaption of the developed systems to changed conditions. Another considered aspect is the economical use of resources to allow an integration of context prediction into embedded systems. The developed architecture is being implemented in terms of a flexible software framework and evaluated with recorded real-world data from everyday situations. This examination shows that the prediction of abstract contexts is already possible within certain limits, but that there is still room for future improvements of the prediction quality.