Decision Making in Ambient Assisted Living Environments based on Uncertain and Fuzzy Data

This Thesis analyses the application of data-mining techniques on fuzzy data in an AAL context. The goal was to research decision making and recommender systems as part of the Reasoning subsystem of an AAL service. It tries to relate information and deduce valid behavioural patterns. The first part of the thesis explores fuzzy clustering and fuzzy control as a mean to describe human behaviour. After a theoretical introduction to clustering and fuzzy control, the Adaptive Online Fuzzy Inference System (AOFIS) is presented and analysed. AOFIS tries to learn from human behaviour and imitate it after a certain learning period. It is based on a doubleclustering technique, which extracts interpretable fuzzy granules. This granules are used to build an if-then-Rulebase as part of an inference system as described in the fuzzy control chapter. The second part of the thesis presents theoretical foundations of recommender systems and several popular techniques to process large amounts of data. It is a necessity to reduce the runtime of such a system and thus improve the performance. Furthermore a fuzzy recommender system was implemented for use as a module in the YouDo it! we help platform as a way to present the user similar videos and topics to the already watched or rated content. This recommender system is a hybrid system and makes use of fuzzy clustering methods to combine the advantages of content-based and collaborative filtering approaches. INHALTSVERZEICHNIS IX