A link prediction algorithm based on convolutional neural network

Link prediction is a fundamental problem in complexity science, focusing on forecasting the emergence of new links or identifying missing links within a given network. In this paper, we propose a novel link prediction method named Link Prediction Based on Convolutional Neural Networks (LPCNN). The approach introduces an innovative feature construction technique and leverages the LeNet-LP convolutional neural network architecture, specifically tailored for link prediction tasks. To assess the performance of LPCNN, extensive experiments were conducted on four publicly available datasets. The experimental results demonstrate that the proposed method significantly enhances link prediction accuracy, highlighting its effectiveness and practical applicability.