Online social networks are highly dynamic and sparse. One of the main problems that arise when analyzing these networks is the problem of predicting the existence of links between users on these networks. This problem is called Link Prediction problem. One of the methods used to deal with the link prediction task in social networks is the supervised learning approach by learning a binary classifier to predict whether or not there exists a link between pairs of actors in these networks. This approach is based on either graph features (such as neighborhood overlap) or node features (such as interests). Many researches have been conducted for studying the link prediction using a variety of techniques like the decision tree and the logistic regression approaches. Some of them rely only on graph-based features, while others combine both the graph and node features. In this thesis, a Rough Set based classification approach was adopted in predicting links over the Facebook social network based on homophilic features. Moreover, other different classification algorithms such as Naïve Bayes, J48 decision tree, Support Vector Machine, Logistic Regression and Multilayer Perceptron Neural Network were employed and compared with the Rough Set based classifier. In addition, the effect of adding two more attributes to the set of homophilic features was studied: the Common Groups and Page Likes. They are used in order to test their effect on the link prediction. Moreover, we compared the effect of combining both homophilic features with some topological features. Two different evaluation techniques were applied: Holdout technique and 10-cross Fold validation technique. F1 measure and Area under the Curve (AUC) were used as performance metrics to compare the performance of the classifiers employed. Our results showed that most of these classifiers can predict link existence with good performance, however the rough set based classifier outperformed all the other classifiers in the case of using homophilic features exclusively and when combining both homophilic and topological features.