Abstract
Web Services are the leading technology for defining and sharing applications and functionalities across the Internet. The use of Web services as the basic construct of Service-Oriented Computing is now, also spreading within dynamic environments such as Pervasive Computing, Cloud Computing, Wireless Sensor Networks, Internet-of-Things (IoT), and Web of Things (WoT). In dynamic service environments, the service conditions might change frequently and sometimes a service might disappear for various reasons (e. g. , low battery or network breakage). Investigation of related works in the literature reveals that the adaptation of service-oriented technologies (such service publication, discovery, and composition) for dynamic service environments has not been sufficiently handled. In order to provide service publication, discovery, and composition solutions of the same accuracy and reliability as the ones in static environments, service discovery needs to take a more automated and knowledge-driven approach. In this thesis we set up the basis for automated service discovery and composition by using probabilistic latent factor models to create a homogeneous indexing and clustering scheme. The probabilistic clustering was compared to commonly used clustering techniques and the method produced more efficient clustering schemes than the other techniques. We proposed a hybrid semantic service matchmaking solution for automated service discovery. The solution consists of a probabilistic matchmaking mechanism and a logic-based matchmaking mechanism. The hybrid matchmaker performs better than state of the art semantic service matchmakers in terms of finding the most relevant services to a client’s service request. Automated service composition is achieved by using a Divide and Conquer algorithm that dynamically creates the execution plan for composition in real-time. The algorithm can also be used for real-time service compensation, making it possible to adapt to the changes in the dynamic service environment. Evaluation results show that our proposed method performs effective service composition and compensation even where a Back-Chaining algorithm fails.