An Empirical Study of the GraphSAGE and Word2vec Algorithms for Graph Multiclass Classification

Abstract

A fundamental task in machine learning involves node classification. However, when considering the context of large graph data, this problem becomes much more challenging. In this paper, we use the Wikipedia hyperlink dataset to evaluate our semi-supervised node classification model. Given a small set of labeled nodes, we develop a multiclass classifier that utilizes the network structure as well as textual descriptions of nodes to predict the most probable category(label) for each test node in a semi-supervised setting. Our experiment shows promising results for graph multiclass classification using directed graphSAGE and word2vec algorithms together. We also visualize the node embeddings in 2D using the t-SNE method.