#引用

##LaTex

@ARTICLE{6381531, author={B. Xue and M. Zhang and W. N. Browne}, journal={IEEE Transactions on Cybernetics}, title={Particle Swarm Optimization for Feature Selection in Classification: A Multi-Objective Approach}, year={2013}, volume={43}, number={6}, pages={1656-1671}, keywords={Pareto optimisation;evolutionary computation;feature extraction;particle swarm optimisation;pattern classification;sorting;PSO-based multiobjective feature selection algorithms;Pareto front;classification performance;classification problems;crowding;dominance;evolutionary multiobjective algorithms;feature subsets;multiobjective particle swarm optimization;mutation;nondominated solutions;nondominated sorting;single objective feature selection method;two-stage feature selection algorithm;Error analysis;Heuristic algorithms;Optimization;Search problems;Standards;Support vector machines;Training;Feature selection;multi-objective optimization;particle swarm optimization (PSO);0}, doi={10.1109/TSMCB.2012.2227469}, ISSN={2168-2267}, month={Dec},}

##Normal

B. Xue, M. Zhang and W. N. Browne, “Particle Swarm Optimization for Feature Selection in Classification: A Multi-Objective Approach,” in IEEE Transactions on Cybernetics, vol. 43, no. 6, pp. 1656-1671, Dec. 2013. doi: 10.1109/TSMCB.2012.2227469 keywords: {Pareto optimisation;evolutionary computation;feature extraction;particle swarm optimisation;pattern classification;sorting;PSO-based multiobjective feature selection algorithms;Pareto front;classification performance;classification problems;crowding;dominance;evolutionary multiobjective algorithms;feature subsets;multiobjective particle swarm optimization;mutation;nondominated solutions;nondominated sorting;single objective feature selection method;two-stage feature selection algorithm;Error analysis;Heuristic algorithms;Optimization;Search problems;Standards;Support vector machines;Training;Feature selection;multi-objective optimization;particle swarm optimization (PSO);0}, URL:

#摘要

Irrelevant and redundant features

Feature selection

a small number of relevant features

• maximizing the classification performance
• minimizing the number of features

the first study on multi-objective particle swarm optimization (PSO) for feature selection

• nondominated sorting
• crowding, mutation, and dominance

compared with：

• two conventional feature selection methods
• a single objective feature selection method
• a two-stage feature selection algorithm
• three well-known evolutionary multi-objective algorithms

12 benchmark data sets

#主要內容

##PSO

##FS

Traditional：

• SFS
• SBS
• the “plus-l-take away-r” method
• sequential forward floating selection (SFFS)
• sequential backward floating selection (SBFS)
• The Relief algorithm — assigns a weight to each feature
• Decision trees (DTs)
• The FOCUS algorithm — exhaustively

##演算法

TP，TN，FP，FN — 真陽性、真陰性、假陽性和假陰性

ER — the error rate obtained by using all available features for classification on the training set

###NSPSOFS

###CMDPSOFS

###試驗