基于人工蜂群优化的帕累托前沿特征选择

#引用

##LaTex

@article{HANCER2018462,
title = “Pareto front feature selection based on artificial bee colony optimization”,
journal = “Information Sciences”,
volume = “422”,
pages = “462 - 479”,
year = “2018”,
issn = “0020-0255”,
doi = “https://doi.org/10.1016/j.ins.2017.09.028”,
url = “http://www.sciencedirect.com/science/article/pii/S0020025516312609”,
author = “Emrah Hancer and Bing Xue and Mengjie Zhang and Dervis Karaboga and Bahriye Akay”,
keywords = “Feature selection, Classification, Multi-objective optimization, Artificial bee colony”
}

##Normal

Emrah Hancer, Bing Xue, Mengjie Zhang, Dervis Karaboga, Bahriye Akay,
Pareto front feature selection based on artificial bee colony optimization,
Information Sciences,
Volume 422,
2018,
Pages 462-479,
ISSN 0020-0255,
https://doi.org/10.1016/j.ins.2017.09.028.
(http://www.sciencedirect.com/science/article/pii/S0020025516312609)
Keywords: Feature selection; Classification; Multi-objective optimization; Artificial bee colony

---

#摘要

to maximize the classification performance
to minimize the number of selected features

the curse of dimensionality

a multi-objective problem

multi-objective artificial bee colony algorithm integrated with non-dominated sorting procedure and genetic operators

ABC with binary representation and ABC with continuous representation

12 benchmark datasets

linear forward selection, greedy stepwise backward selection, two single objective ABC algorithms and three well-known multi-objective evolutionary computation algorithms

---

#主要内容

ABC

---

##Non-EC approaches

wrapper：

• sequential forward selection (SFS)
• sequential backward selection (SBS)
• sequential forward floating selection (SFFS)
• sequential backward floating selection (SFBS)

filter：

• FOCUS — exhaustive search
• Relief
• information theoretic approaches： MIFS, mRmR and MIFS-U

---

##提出的算法

objectives:

• (1) minimizing the feature subset size
• (2) maximizing the classification accuracy

multi-objective ABC based wrapper approach

Bin-MOABC and Num-MOABC

###A calculate probabilities for onlookers:

###B update individuals:

greedy selection

a fast non-dominated sorting scheme

###C generate new individuals:

binary solution generator (BSG)：

numeric solution generator (NSG))：

###D Representation and fitness function:

activation codes — 【0，1】

vary / discrete

a user specified threshold value — 0.5、

The classification error rate of a feature subset：

---

##试验

UCI machine learning repository

加权和

ER — the error rate calculated through all available features in the dataset

NSGAII, NSSABC and multi-objective PSO (MOPSO)