The day mankind met with smart-phones, a new era started. Since then, daily mobile internet usage rates are increasing everyday and people have developed new habits like frequently sharing information (photo, video, location, etc.) on online social networks. Location Based Social Networks (LBSNs) are the platforms that empowers users to share place/location information with friends. As all other social networks, LBSNs aim to acquire more users with a smart friend recommendation. Solution for smart friend recommendation problem is studied under link prediction field by researchers. Check-in information is the main data for link prediction in LBSNs. Data extracted from check-in information plays vital role for predictor performance. In this study, we attempt to make use of detailed analysis of place category in order to exploit possible information gain enhancements through such semantic information. We proposed two new feature groups; Common Place Check-in Count Product Sum and Common Category Check-in Count Sum Product. For any link candidate pair; those features are calculated for each category. Use of new features improved the link prediction performance for multiple data subsets.