Hybrid ensemble approaches to online harassment detection in highly imbalanced data

作者：

Highlights：

• Research addressed most popular issues of NLP: cyberharassment and class imbalance.

• Nine recent and widely approaches are used for imbalance learning task.

• Several learning models examined in various scenarios to find top performing ones.

• Exploration of hybrid approaches based on deep learning has produced good results.

摘要

•Research addressed most popular issues of NLP: cyberharassment and class imbalance.•Nine recent and widely approaches are used for imbalance learning task.•Several learning models examined in various scenarios to find top performing ones.•Exploration of hybrid approaches based on deep learning has produced good results.

论文关键词：Online harassment detection,Imbalanced learning,Word embedding,Deep learning,Imbalanced Twitter data

论文评审过程：Received 27 October 2020, Revised 10 January 2021, Accepted 16 February 2021, Available online 25 February 2021, Version of Record 23 March 2021.

论文官网地址：https://doi.org/10.1016/j.eswa.2021.114751