A Nearest Prototype Selection Algorithm Using Multi-objective Optimization and Partition

被引:3
作者
Li, Juan [1 ,2 ]
Wang, Yuping [1 ]
机构
[1] Xidian Univ, Sch Comp Sci & Technol, Xian 710071, Peoples R China
[2] Shaanxi Normal Univ, Sch Distance Educ, Xian, Peoples R China
来源
2013 9TH INTERNATIONAL CONFERENCE ON COMPUTATIONAL INTELLIGENCE AND SECURITY (CIS) | 2013年
关键词
machine learning; prototype selection; multi-objective optimization; genetic algorithm; divide-and-conquer partition;
D O I
10.1109/CIS.2013.62
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Prototype selection aims at reducing the scale of datasets to improve prediction accuracy and operation efficiency by removing noisy or redundant patterns via the nearest neighbor classification algorithms. Genetic algorithms have been used recently for prototype selection and showed good performance, however, they have some drawbacks such as the deteriorated running effect, slow convergence for the large datasets. The goal of designing good prototype selection algorithm is to maximize prediction classification accuracy and minimize the reduction ratio simultaneously. According to this goal, a two objective optimization model is set up for prototype selection problem. To make the model easier to be solved, the model is transformed to a single objective optimization model by the division of the two objectives. To solve the problem efficiently, a new prototype selection algorithm based on genetic algorithm and the divide-and-conquer partition is presented. The divide-and-conquer partition can divide the whole dataset into some random subsets, and thus make the problems more easier. Then, the genetic algorithm is specifically designed on these divided random subsets. The simulation results indicate that the proposed algorithm can obtain smaller reduction ratio and higher classification efficiency, or at least comparable to those of some existing compared algorithms. This illustrates that the proposed algorithm is an expedient method in design nearest neighbor classifiers.
引用
收藏
页码:264 / 268
页数:5
相关论文
共 15 条
[1]  
Calana Yenisel Plasencia, 2010, Proceedings of the 2010 20th International Conference on Pattern Recognition (ICPR 2010), P177, DOI 10.1109/ICPR.2010.52
[2]   Stratification for scaling up evolutionary prototype selection [J].
Cano, JR ;
Herrera, F ;
Lozano, M .
PATTERN RECOGNITION LETTERS, 2005, 26 (07) :953-963
[3]   Using evolutionary algorithms as instance selection for data reduction in KDD: An experimental study [J].
Cano, JR ;
Herrera, F ;
Lozano, M .
IEEE TRANSACTIONS ON EVOLUTIONARY COMPUTATION, 2003, 7 (06) :561-575
[4]  
Chang E.I., 1991, Advances in neural information processing systems, V3, P797
[5]  
Eshelman LJ., 1991, FDN GENETIC ALGORITH, P265, DOI DOI 10.1016/B978-0-08-050684-5.50020-3
[6]   A memetic algorithm for evolutionary prototype selection:: A scaling up approach [J].
Garcia, Salvador ;
Cano, Jose Ramon ;
Herrera, Francisco .
PATTERN RECOGNITION, 2008, 41 (08) :2693-2709
[7]  
Goldberg DE., 1989, GENETIC ALGORITHMS S, V13
[8]   CONDENSED NEAREST NEIGHBOR RULE [J].
HART, PE .
IEEE TRANSACTIONS ON INFORMATION THEORY, 1968, 14 (03) :515-+
[9]   Design of an optimal nearest neighbor classifier using an intelligent genetic algorithm [J].
Ho, SY ;
Liu, CC ;
Liu, S .
PATTERN RECOGNITION LETTERS, 2002, 23 (13) :1495-1503
[10]   Comparison of genetic algorithm based prototype selection schemes [J].
Babu, T. Ravindra ;
Murty, M. Narasimha .
Pattern Recognition, 2001, 34 (02) :523-525