Bachelor Thesis BCLR-2016-49

BibliographySchmalfuss, Jenny: Simulation and Adaptation of Contextual Bandit Algorithms for IoT Service Discovery.
University of Stuttgart, Faculty of Computer Science, Electrical Engineering, and Information Technology, Bachelor Thesis (2016).
137 pages, english.
CR-SchemaI.2.6 (Artificial Intelligence Learning)
Abstract

As the emerging Internet of Things (IoT) makes an increasing ammout of connected services accessible to a broad range of users, the identification of contextually relevant services becomes an indispensable task. In order to evaluate existing methods for contextual IoT service recommendation, we develop an ambient space simulation to emulate huge numbers of IoT services and user interactions. Secondly we investigate a new extension of the previously evaluated single IoT service recommendation - the recommendation of composite services consisting of multiple services to perform a mutual task. To address this challenge, we construct and implement a framework called ConComM to identify services that are likely to work well together for a joint task. Our previously developed ambient space simulation is then used to evaluate our frameworks performance. The framework itself utilizes a novel k-cut algorithm based on a modification of the existing k-cut procedure SPLIT. We call this new procedure SPLITrel and show it to outperform all tested benchmark algorithms for minimum k-cuts on graphs used by ConComM. Our experiments prove ConComM to significantly push the performance of an existing single service recommendation approach for composite service recommendation. Within this work we do not only provide a simulation that allows to evaluate recommendation systems in environments densly filled with IoT services. We also develop a framework that enables contextual bandit algorithms to provide improved recommendations for composite services. 3

Full text and
other links
PDF (36210566 Bytes)
Access to students' publications restricted to the faculty due to current privacy regulations
Department(s)University of Stuttgart, Institute of Visualisation and Interactive Systems, Visualisation and Interactive Systems
Superviser(s)Schmidt, Prof. Albrecht; Dingler, Tilman; Carlson, Dr. Darren
Entry dateSeptember 26, 2018
   Publ. Computer Science