To develop practical Ubiquitous Computing (UbiComp) applications, we design, formalize and implement a Proactive Service Mechanism (PSM). However, in UbiComp environments, imprecise and incomplete context information, as well as limited resource, introduces novel challenging issues for PSM from service model and Quality of Service (QoS) model to service selection algorithm. In our design, we propose a user-oriented QoS model and a componentbased service model to support adaptive service selection. In particular, we design fuzzy linguistic terms to help the user express his/her preference to quality criteria with natural language. Furthermore, the user?s preference can be mapped into the weights of quality criteria by employing fuzzy set theory. The kernel of the service selection problem is identified as Multiple Attribute Decision Making (MADM) problem with incomplete preference, which can be solved by classic methods. Finally, we conduct several experiments to testify the performance and scalability of our design.