Based on large scale click-through data, this paper study the interactive process between user and search engine, and derive user decision process. A comparative study between clicks on relevant results and non-relevant ones analyzes the reliability of individual user click-through behavior on search results. Three types of features are proposed and estimated for separating reliable user clicks from other ones. Experimental results show that the proposed method evaluates the reliability of user behaviors effectively based on click context features of Web search users.