Title: Bagging method-based classifier chains for multi-label classification

Authors: Jiaman Ding; Haotian Tan; Runxin Li; Yuanyuan Wang; Lianyin Jia

Addresses: Faculty of Information Engineering and Automation, Kunming University of Science and Technology, China; Artificial Intelligence Key Laboratory of Yunnan Province, Kunming, 650500, China ' Faculty of Information Engineering and Automation, Kunming University of Science and Technology, China; Artificial Intelligence Key Laboratory of Yunnan Province, Kunming, 650500, China ' Faculty of Information Engineering and Automation, Kunming University of Science and Technology, China; Artificial Intelligence Key Laboratory of Yunnan Province, Kunming, 650500, China ' Australian National University, Canberra, 650500, Australia ' Faculty of Information Engineering and Automation, Kunming University of Science and Technology, China; Artificial Intelligence Key Laboratory of Yunnan Province, Kunming, 650500, China

Abstract: Classifier chains are one of the main methods for dealing with multi-label classification. For classifier chains, it is important to order labels according to the correlations among labels and find the corresponding feature subsets. Therefore, we propose a bagging method-based classifier chain for multi-label (BMCC4ML). Firstly, a unique feature subset for each label is selected by measuring the feature importance of labels based on ReliefF. Next, considering the positive and negative correlations from a global perspective, all the labels are in order by calculating the correlations among labels. After that, inspired by ensemble learning methods, the ensemble classifier chains were constructed according to the strategy that the number of chains and the repeat rate of label order were higher than given thresholds to improve the prediction accuracy. Experiments on publicly accessible multi-label datasets demonstrate that BMCC4ML achieves more prominent results than other related approaches across various evaluation metrics.

Keywords: multi-label classification; ensemble chains; ReliefF; feature selection; bagging method.

DOI: 10.1504/IJCSM.2022.127831

International Journal of Computing Science and Mathematics, 2022 Vol.16 No.2, pp.112 - 124

Received: 22 Oct 2021
Accepted: 17 Mar 2022
Published online: 19 Dec 2022 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article