Paper Title
Inter Common Neighbor Connections in Supervised Link Prediction

Most social, biology, and information systems can use a network to define, therefore, examination of complex networks has become an important branch of many scientific fields. Guess the link to predict whether there will be a connection between the two nodes based on the relevant information and the available information of the link is the Link Prediction Problem. Recommender systems are the most common examples where link prediction is used extensively. The main approach to solve the Link Prediction Problem is using Supervised Learning; a score based approach where features from nodes of current snapshot of a graph are extracted and based on that future links are predicted. In this paper, we propose a supervised learning link prediction model using inter-common neighbor connection strength as a feature that considers how well the common neighbors of a node pair are connected alongside other commonly used graph topology based features to predict future links in social network based datasets. We compare how using inter-common neighbor connection strength improves the predictions using AUC-ROC score and F1 score as metric. Keywords - Link Prediction, Recommender Systems, Supervised Learning, AUC-ROC score, F1 score