Social Networks are among the emerging services in the current virtual world which are used as a connecting medium between users over the Internet. Users generally connect to social networks to interact with their family members, coworkers, or even for commercial goals. Hence, the key for the success of every social network provider lies on satisfying user interest and thereby, expanding the network. Accordingly, one way to increase users' interest toward a social network is introducing them to other users with the similar characteristics using recommender systems. The current article introduces a recommender system based on AHP method and semantic similarities as well as color psychology between users in an online social network. To evaluate the proposed system, we used 5705 real Twitter celebrity profiles as alternative for recommendation, together with 100 virtual profiles as ordinary users. The experimental results indicate maximum recommendations error rate as low as 12.5.