With the tremendous increase of web services published online, the problem of selecting the best service offers becomes more challenging. Users need to make their decisions on multiple and conflicting non-functional requirements. It is a natural fit to apply the Multi-Criteria Decision Making (MCDM) theory to the service selection and ranking process. In our proposed QoS-based service selection system, we take the user-centric standpoint to design the system. We improve the original MCDM models so that the user requirements on the QoS criteria are included in the rank calculation process. Our proposed QoS weighting method considers the well-defined ANP method combined with the user-defined weights. We compared the improved selection methods and we found that the Constraint Programming method is the best in terms of its sensitivity to the changes made to the QoS weights. Consequently, the results produced from this comparison would be presented to the user.