.png){kind=icon}
返回
A decomposition-based multi-objective immune algorithm for feature selection in learning to rank
.png){kind=icon}
PRE
AI
DOI：10.1016/j.knosys.2021.107577.png)
.png){kind=icon}
原文求助.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
Learning-to-rank (L2R) based on feature selection has been proved effectively. However, feature selection problem is more challenging due to two conflicting objectives, namely, maximizing classification performance and minimizing the number of features. Therefore, in this article, a novel decomposition-based multi-objective immune algorithm for feature selection in L2R, called MOIA/DFSRank, is proposed. The proposed algorithm associates each solution with a scalar subproblem based on Tchebycheff decomposition approach, which makes the optimization process more efficient. In addition, we propose an Elite Selection Strategy (ESS) in initialization phase, which can significantly improve the diversity and convergence of the initial population. Moreover, the proposed algorithm uses two effective operators, a clonal selection operator and a mutation operator, where the clonal selection operator is to generate clone population to better guide the search direction of the evolution and the mutation operator aims to retain excellent features with a higher probability in evolution. In this paper, the pairwise method with O(n2) size is used to train the ranking model, and extensive experiments are conducted on the four public LETOR benchmark data sets. The experimental results demonstrate that the proposed algorithm can obtain significant performances on the ranking accuracy and the number of features. (c) 2021 Elsevier B.V. All rights reserved.
Keyword:
Learning-to-rank
Multi-objective optimization
Immune algorithm
Decomposition
Large-scale data
期刊
K
IF:
7.6
论文数:
1.3W
被引数:
4.5W
机构
引用论文
A PSO-based multi-objective multilabel feature selection method in classification.png)
SCIENTIFIC REPORTS
OAAI
SCIENTIFIC REPORTSIF3.9
OAAI分享收藏Fast wrapper feature subset selection in high-dimensional datasets by means of filter re-ranking
PREAI
分享收藏Feature weighting and selection with a Pareto-optimal trade-off between relevancy and redundancy
PREAI
分享收藏
查看更多内容