Probabilistic Estimation Quality Ratings of Online Services
Accurate estimation of quality of online services is both an important and difficult problem, since a service has many interdependent quality attributes influenced by several contextual factors. It is even more challenging as quality ratings come from sources with unknown reliability, each source may rate a service on different quality aspects. Although several solutions have been proposed, there is little work addressing all these issues thoroughly. In this paper, we show that domain knowledge on service structure and related constraints, such as causal dependencies among quality attributes and contextual factors, while widely available, can be exploited to effectively address the above issues in a theoretically-sound framework. Theoretical analysis shows that computational cost of the approach is acceptable, and accurate evaluation of service quality requires a reasonable number of user feedback, provided services have a small number of quality attributes and contextual factors.