Link Prediction and Node Classification on Citation Network
Chenyu Han, Xiaoyu Fu, Yaohua Liang
Abstract
In this project, we use Graph Convolutional Networks model to do link prediction and node classification on citation network. Then we compare the performance of this model to benchmark methods, including common neighbors, jaccard coefficient, multilayer perceptron and support vector machine. However, GCN can be difficult to interpret and explain. In order to improve the explainability of the model, we add attention mechanism to original GCN and extract top-5 attention weights and their corresponding edges. In this way, we can better understand the model's decision making process.
Topics & Concepts
Computer scienceJaccard indexBenchmark (surveying)Node (physics)Support vector machineArtificial intelligenceMultilayer perceptronMachine learningData miningGraphLink (geometry)Artificial neural networkTheoretical computer sciencePattern recognition (psychology)Structural engineeringGeographyEngineeringComputer networkGeodesyComplex Network Analysis TechniquesAdvanced Graph Neural NetworksBioinformatics and Genomic Networks