Service Discovery and Recommendation in Rough Hierarchical Granular Computing

Service discovery is one of the most vital components involved in almost all service applications. A lot of researches have paid attention to improving the matching accuracy for the given user requirement. Without a clarified requirement specification to describe the goals user really wants to achieve, any matching algorithm is useless. Goal-oriented requirement engineering is a formal requirement analysis methodology which recursively decompose a complex requirement into a set of finer grained goals. Such a hierarchical granulation structure partitions a requirement into a family of fine grain-sized granules. In order to handle uncertainties, rough set theory is employed in granular computing. For any given imprecise user requirement, a set of ordered stratified rough set approximations can be induced over all possible partitions. These approximations are used to iteratively refine imprecise requirement and recommend goals most probably desired. A matching algorithm based on six types of granule approximation is also given to describe different matching strategies for satisfying user requirement. Through the theoretical analysis and case study, it shows that rough set theory combining with granular computing is powerful to handle imprecise requirements and also provide better service quality.