Fusion: Practice and Applications FPA 2692-4048 2770-0070 10.54216/FPA https://www.americaspg.com/journals/show/3448 2018 2018 College of Computer Science and Information Technology, University of Karbala, Karbala, Iraq Raaid Raaid Babylon Education Directorate, Ministry of Education, Babil, Iraq Nisreen Abbas Hussein Al-Ayen Iraqi University, Thi-Qar, Iraq; College of Information Technology, University of Babylon, Babil, Iraq Raaid Alubady Recently, the complex network has become popular use as it can transfer huge amounts of multimedia, text, ideas, and other information, encouraging many participant connections. Social media is one of these networks that make the most connections. Predicting the formation or dissolution of links between nodes presents a problem for social network analysis researchers. Since social networks are dynamic, this task is exciting as it may also forecast lost network links with less information. On the other way, current link prediction methods use simply node similarity to find links. This study proposes a new technique that relies on node attributes and similarity measures. Nodes are labeled by their centrality and similarity. The network's edges are negative and positive samples. A well-defined dataset for link prediction comprises the features of the nodes at the edges labeled either positive or negative. The dataset is passed to multiple machine learning classifiers. On several real-world networks. The experiments conducted during the research show that Gradient Boosting gave the highest accuracy of 99% compared with other methods. 2025 2025 41 55 10.54216/FPA.180104 https://www.americaspg.com/articleinfo/3/show/3448