Multi-objective Service Query Optimization
Existing service optimization approaches usually select services based on a predefined objective function [59, 94]. They require users to express their preference over different (and sometimes conflicting) quality parameters as numeric weights. The objective function assigns a scalar value to each service provider based on the quality values and the weights given by the service user. The provider gaining the highest value from the objective function will be selected and returned to the user. Implementing such an optimization strategy may pose several challenges: Transforming personal preferences to numeric weights is a rather demand- ing task for users. Sometimes it is even impossible if the preference is still vague before the user is presented with the actual service providers. Users may miss their desired providers because of an imprecise specification of the weights, which would be very common in real-world scenarios. Users may lose the flexibility to select their desired providers by them- selves. For example, a service user may choose a service provider that has a good reputation within a price range she can tolerate although price is a very important factor she considers. In this case, the relationship between reputation and price is subtle and the choice from different users may vary significantly. Therefore, it would be wise to give users the flexibility make their own selections from a small set of candidate providers.
KeywordsHull Convolution Sorting
Unable to display preview. Download preview PDF.