Twitter is one of the largest social media platforms in the world. Although Twitter can be used as a tool for getting valuable information related to a topic of interest, it is a hard task for us to find users to follow for this purpose. In this paper, we present a method for Twitter user recommendation based on user relations and taxonomical analysis. This method first finds some users to follow related to the topic of interest by giving keywords representing the topic, then picks up users who continuously provide related tweets from the user list. In the first phase we rank users based on user relations obtained from tweet behaviour of each user such as retweet and mention (reply), and we create topic taxonomies of each user from tweets posted during different time periods in the second phase. Experimental results show that our method is very effective in recommending users who post tweets related to the topic of interest all the time rather than users who post related tweets just temporarily.