The power of ensemble learning in sentiment analysis
作者：
Highlights：
• Several ensemble techniques are comparatively evaluated in respect of benchmark data.
• Median performance improvements of up to 5.53% over individual models are achieved.
• A novel ensemble selection method is proposed that minimises storage and retraining.
• Clear trends for effective ensemble generation are observed.
摘要
•Several ensemble techniques are comparatively evaluated in respect of benchmark data.•Median performance improvements of up to 5.53% over individual models are achieved.•A novel ensemble selection method is proposed that minimises storage and retraining.•Clear trends for effective ensemble generation are observed.
论文关键词：Ensemble learning,Sentiment analysis,Machine learning,Natural language processing
论文评审过程：Received 10 January 2021, Revised 21 June 2021, Accepted 26 August 2021, Available online 5 September 2021, Version of Record 17 September 2021.
论文官网地址：https://doi.org/10.1016/j.eswa.2021.115819
