A Novel Friend Recommendation System Using Link Prediction in Social Networks

Abstract

Link prediction is a method used to predict the existence of a non-existing links between two entities within a network. However, the growing size of social networks has made conducting link prediction studies more challenging. This chapter proposes a friend recommendation system that employs feature engineering techniques on a given dataset. The feature engineering process involves extracting relevant features such as shortest path, Katz centrality, Jaccard distances, PageRank, and preferential attachments, etc. Random Forest and XGBoost algorithms are then utilized to recommend non-existent connections by suggesting new edges in the graph. By implementing these approaches, the authors aim to improve the accuracy and effectiveness of friend recommendations in the social network graph. By considering both types of edges in the recommendation process, they enhance the performance of the friend recommendation system. This approach allows leveraging the valuable insights within the network graph, resulting in more accurate and reliable recommendations.