Multi-label classification by exploiting label correlations

作者：

Highlights：

• We model two multi-label classification models by respectively exploiting global and local label correlation.

• By introducing the lower and upper approximation, the models consider the uncertainty during the process of classification.

• The level of granularity has effect on the performance of MLRS and MLRS-LC.

• The inclusion degree impact the performance of MLRS and MLRS-LC.

摘要

•We model two multi-label classification models by respectively exploiting global and local label correlation.•By introducing the lower and upper approximation, the models consider the uncertainty during the process of classification.•The level of granularity has effect on the performance of MLRS and MLRS-LC.•The inclusion degree impact the performance of MLRS and MLRS-LC.

论文关键词：Multi-label classification,Rough sets,Uncertainty,Correlation

论文评审过程：Available online 23 October 2013.

论文官网地址：https://doi.org/10.1016/j.eswa.2013.10.030