Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
6864123 | Neurocomputing | 2018 | 19 Pages |
Abstract
Multi-label classification on social network data deals with the problem of labeling nodes in the network (i.e. instances in the data set) with multiple classes. Existing connectivity-based approaches have been used in classification by exploiting the correlations between linked nodes. However, this popular strategy may not always perform well, as it ignores the neighborhood of nodes and the correlations between nodes and class labels. In this paper, we propose a novel multi-label relational classifier which exploits the correlations between nodes and class labels. We first identify similar nodes for each unlabeled node based on local network structure. Then we perform clustering on nodes with known labels. We introduce an aggregated class probability to capture the correlations between nodes and class labels based on the clustering results. Experiments with real-world datasets demonstrate that our proposed method improves classification performance comparing to the existing approaches.
Related Topics
Physical Sciences and Engineering
Computer Science
Artificial Intelligence
Authors
Zhang Zan, Wang Hao, Liu Lin, Li Jiuyong,