A link prediction algorithm based on support vector machine

Abstract

The path-based similarity index algorithm has proven effective in link prediction, with the Local Path (LP) similarity index leveraging second-order path information to enhance accuracy significantly. However, few machine learning-based link prediction algorithms fully utilize higher-order path information beyond the second order. Addressing this gap, this paper proposes a novel link prediction algorithm, termed Link Prediction based on Support Vector Machine, which incorporates the concept of the LP similarity index into feature vector construction, integrating higher-order path information comprehensively. Extensive controlled experiments on four public datasets demonstrate that our algorithm achieves notable performance improvements compared to traditional similarity index-based link prediction algorithms.