Improving Social Network Link Prediction with an Ensemble of Machine Learning Techniques

Abstract

Finding missing connections in social networks are a crucial task that has generated a great deal of interest lately. To solve this issue, there have been several machine learning methods proposed, however the majority of them concentrate on a particular kind of feature or technique. The goal of link prediction is to identify pairs of nodes that will either form a link in the future or not. In order to accurately predict links in social networks, we provide a hybrid machine learning methodology in this research that integrates many features and techniques. In this research, we present an ensemble strategy to enhance the resilience and accuracy of link prediction by utilizing the advantages of many machine learning approaches. Specifically, we combine seven popular methods, namely, logistic regression, support vector machines, random forests, and decision tree, naïve Bayes, k-NN and gradient boosting, and employ them together in a unique ensemble framework. In addition, we employ principal component analysis to lower the feature space's dimensionality and boost the model's computational effectiveness. To evaluate the proposed method, we run experiments on real-world social network dataset facebook. The results demonstrate that our method outperforms the state-of-the-art approaches in terms of F1 Score, accuracy, precision, and recall. All things considered, our method offers a viable way to improve prediction.