Enhancement of DNN-based multilabel classification by grouping labels based on data imbalance and label correlation
作者:
Highlights:
• DNN-based MLC suffers two critical problems: data imbalance and label correlation
• Neural network configurations with grouped labels were developed to enhance MLC
• Strategies for grouping labels were proposed to tackle two critical problems in MLC
• Experiments show that the proposed method increase accuracy of minority labels
• Adjusting dependence in grouped labels improve accuracy of correlated labels
摘要
•DNN-based MLC suffers two critical problems: data imbalance and label correlation•Neural network configurations with grouped labels were developed to enhance MLC•Strategies for grouping labels were proposed to tackle two critical problems in MLC•Experiments show that the proposed method increase accuracy of minority labels•Adjusting dependence in grouped labels improve accuracy of correlated labels
论文关键词:Multilabel classification,data imbalance,label correlation,neural network
论文评审过程:Received 2 April 2022, Revised 5 August 2022, Accepted 9 August 2022, Available online 12 August 2022, Version of Record 18 August 2022.
论文官网地址:https://doi.org/10.1016/j.patcog.2022.108964