.png){kind=icon}
.png){kind=icon}
Return
ELM: a novel ensemble learning method for multi-target regression and multi-label classification problems
.png){kind=icon}
PRE
AI
DOI：10.1007/s10489-024-05570-3.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
In this paper, a new Ensemble Learning Method (ELM) is proposed to deal with multi-target regression and multi-label classification problems. In ELM, the output of each regressor or classifier unit is ensembled to the final multi-target regression or multi-label classification stage by utilizing the threshold strategy. In the final learning stage, a sorted-by-score scheme is specially tailored to generate the regression or classification results. The proposed method is numerically implemented and applied to twenty-nine benchmark data-sets. The average relative root mean squared error is employed to quantify the accuracy of multi-target regression methods. Meanwhile, metrics such as the hamming loss, the macro-precision, the macro-recall, the macro-F1, the micro-precision, and the micro-F1 are utilized to evaluate multi-label classification methods. Furthermore, the Bonferroni-Dunn post-hoc test is employed to illustrate the relative performance of existing methods. Overall, the corresponding regression or classification results show that the proposed ELM generally quantitatively outperforms the other nineteen existing multi-target regression or multi-label classification methods.
Keywords:
Ensemble learning method
Multi-target regression
Multi-label classification
Threshold strategy