Graph-Based Feature Selection in Classification: Structure and Node Dynamic Mechanisms

特征选择 图形 特征(语言学) 计算机科学 模式识别(心理学) 人工智能 算法 数据挖掘 理论计算机科学 语言学 哲学
作者
Fan Cheng,Changjun Zhou,Xudong Liu,Qijun Wang,Jianfeng Qiu,Lei Zhang
出处
期刊:IEEE transactions on emerging topics in computational intelligence [Institute of Electrical and Electronics Engineers]
卷期号:7 (4): 1314-1328 被引量:4
标识
DOI:10.1109/tetci.2022.3225550
摘要

Recently, researchers pay more attention to designing graph-based methods to address the feature selection problem, since these methods can effectively utilize the underlying topology structure and complex relationships between nodes in the constructed feature graph. Therefore, they can obtain the feature subset with high quality. The existing graph-based methods mainly focus on using different graph-theoretical techniques to select features from the constructed feature graphs. However, little attention is focused on constructing a suitable feature graph for feature selection, which is also an important component for achieving a good feature subset. To fill the gap, in this paper, a novel graph-based algorithm named GBFS-SND is proposed for feature selection, where the structure and node dynamic mechanisms are designed to directly optimize the performance of feature selection. To be specific, in GBFS-SND, a candidate feature graph is firstly created by considering both the importance of feature and the relations between features. Then, on the created candidate graph, an MOEA-based structure dynamic mechanism is suggested to acquire a feature subgraph with better structure, from which we can obtain a promising feature subset. Finally, a node dynamic mechanism is also suggested, with which the weights of the nodes are dynamically adjusted as the structure of feature graph changes. Thus, the performance of GBFS-SND can be further enhanced. Empirical studies are conducted by comparing the proposed algorithm with several state-of-the-art feature selection methods on different data sets. The experimental results demonstrate the superiority of GBFS-SND over the comparison methods in terms of both the accuracy and the number of selected features.
最长约 10秒,即可获得该文献文件

科研通智能强力驱动
Strongly Powered by AbleSci AI
科研通是完全免费的文献互助平台,具备全网最快的应助速度,最高的求助完成率。 对每一个文献求助,科研通都将尽心尽力,给求助人一个满意的交代。
实时播报
追寻从灵完成签到,获得积分10
刚刚
追寻若南发布了新的文献求助10
刚刚
1秒前
韩较瘦完成签到,获得积分0
1秒前
麦当当完成签到 ,获得积分10
1秒前
大个的应助被谨慎的秋烟采纳,获得10
2秒前
xxxx完成签到,获得积分10
2秒前
彭于晏的应助被miuwu采纳,获得10
2秒前
3秒前
情怀的应助被虎杖悠仁采纳,获得10
3秒前
飘逸子轩完成签到,获得积分10
4秒前
null的应助被mango采纳,获得10
4秒前
orixero的应助被负责以山采纳,获得10
4秒前
4秒前
申晨曦发布了新的文献求助10
4秒前
xxxx发布了新的文献求助10
5秒前
6秒前
我是老大的应助被清爽秋白采纳,获得10
6秒前
7秒前
Ava的应助被欣喜巧曼采纳,获得10
8秒前
专注的荣轩完成签到,获得积分10
8秒前
8秒前
彭于晏的应助被nmn采纳,获得10
8秒前
8秒前
Tuniverse_发布了新的文献求助10
9秒前
hongfangpan完成签到,获得积分10
9秒前
9秒前
10秒前
10秒前
万能图书馆的应助被大聪明采纳,获得10
11秒前
11秒前
xiaoyu完成签到 ,获得积分10
11秒前
跑不掉的可乐猪完成签到,获得积分10
11秒前
11秒前
霍则风发布了新的文献求助10
11秒前
vkingda完成签到,获得积分10
12秒前
SciGPT的应助被hope采纳,获得30
12秒前
霍则风发布了新的文献求助10
13秒前
13秒前
科研通AI6.4的应助被小困包采纳,获得10
14秒前
高分求助中
(应助此贴封号)【重要!!请各用户(尤其是新用户)详细阅读】【科研通的精品贴汇总】 10000
拉伸高分子熔体中缠结行为的模拟和小角中子散射研究 1700
The World Atlas of Wine 9th edition 1000
奈特基础组织学:配套组织病理学(第4版)(Netter's Essential Histology With Correlated Histopathology) 800
John McDowells Bildungstheorie 800
Markov Decision Processes and Reinforcement Learning 700
八旗制度の研究 600
热门求助领域 (近24小时)
化学 材料科学 医学 生物 纳米技术 计算机科学 工程类 有机化学 化学工程 生物化学 物理 复合材料 内科学 催化作用 光电子学 人工智能 细胞生物学 无机化学 物理化学 数学
热门帖子
关注 科研通微信公众号,转发送积分 8171039
求助须知:如何正确求助?哪些是违规求助? 9563751
关于积分的说明 21772122
捐赠科研通 7762548
什么是DOI,文献DOI怎么找? 3430786
关于科研通互助平台的介绍 2554676
邀请新用户注册赠送积分活动 2475657