Optimization and label propagation in bipartite heterogeneous networks to improve transductive classification of texts
作者：
Highlights：
•
• Scalable algorithm based on bipartite networks to perform transduction.
• Unlabeled data effectively employed to improve classification performance.
• Better performance than algorithms based on vector space model or networks.
• Rigorous evaluation to show the drawbacks of the existing transductive algorithms.
• Trade-off analysis between inductive supervised and transductive classification.
摘要
•Scalable algorithm based on bipartite networks to perform transduction.•Unlabeled data effectively employed to improve classification performance.•Better performance than algorithms based on vector space model or networks.•Rigorous evaluation to show the drawbacks of the existing transductive algorithms.•Trade-off analysis between inductive supervised and transductive classification.
论文关键词：Text classification,Transductive learning,Graph-based learning,Text mining,Label propagation,Bipartite heterogeneous network
论文评审过程：Received 28 July 2014, Revised 22 April 2015, Accepted 6 July 2015, Available online 6 November 2015, Version of Record 19 February 2016.
论文官网地址：https://doi.org/10.1016/j.ipm.2015.07.004
