Exploring Efficiency of Data Mining Techniques for Missing Link in Online Social Network

Missing link in Online Social Network (OSN) is an interesting problem for capturing missing relation and understanding user's behavior. The existing work introduced social features for training predictive models, but they used only SVM prediction technique for solving the problem. However, we suspect that other prediction techniques may give better performance. This study investigates prediction performances of SVM, k-NN, Decision Tree, Neural Networks, Naïve Bayes, Logistic Regression and Random Forest using two OSN datasets (high-density and low density). We realize that the Random Forest technique has the best performance with F1-measure score. Moreover, this technique is most robust technique for the both datasets.