一种新的人工免疫网络算法及其在复杂数据分类中的应用

刘若辰; 钮满春; 焦李成

doi:10.3724/SP.J.1146.2009.00309

一种新的人工免疫网络算法及其在复杂数据分类中的应用

doi: 10.3724/SP.J.1146.2009.00309

计量
- 文章访问数: 3601
- HTML全文浏览量: 78
- PDF下载量: 1386
- 被引次数: 0
出版历程
- 收稿日期: 2009-03-12
- 修回日期: 2009-10-09
- 刊出日期: 2010-03-19

A New Artificial Immune Network Algorithm for Classifying Complex Data

摘要

摘要: 作为一种新的智能计算方法，人工免疫网络已被广泛的应用到模式识别以及数据分类中。现有的人工免疫网络分类算法大都存在两个缺陷：一是网络规模庞大、计算复杂；二是对抗原的一次递呈并不能保证获得全局最优分类器。该文提出了一种新的人工免疫网络分类算法，该算法利用每个类别对应单个B细胞的策略，简化网络规模并减少了同类别B细胞之间的抑制操作，同时引入了新的基于对训练样本正确识别率的亲合度评价函数，实现了基于抗原的优先级的选择策略。采用5组UCI的线性数据和4组混合特征数据以及1幅SAR图像对算法的性能进行了全面测试，结果表明，与模糊C均值算法，多值免疫(MVIN)算法和基于分类问题的克隆选择算法(CSA)相比，新算法在分类精度上具有一定的优势，鲁棒性更好。
- 人工免疫网络; 分类; SAR图像; 混合数据
Abstract: As a new computational intelligence method, the Artificial Immune Network (AIN) is widely applied to pattern recognition and data classification. Existing artificial immune network algorithms for classifier have two major limitations: one is the scale of the networks, a large scale of networks needs high computation complexity, the other is only once presenting the antigens that can not guarantee find the optimal global classifier. A new Artificial Immune Network Classifier (AINC) algorithm is proposed in this paper. In the proposed algorithm, only one B-cell is used to denote single class in order to reduce the scale of network, and avoid the suppression operation between B-cells, moreover, a new affinity based on the correct rate is proposed to realize the evaluation strategy based on antigen priority. The proposed algorithm is extensively compared with Fuzzy C-Means (FCM), Multiple-Valued Immune Network algorithm (MVIN), and Clonal Selection Algorithm for classifier (CSA) over a test suit of several real life data sets and one SAR image. The result of experiment indicates the superiority of the AINC over FCM, MVIN and CSA on accuracy and robustness.

HTML全文

参考文献(1)

Hunt J E and Cooke D E. Learning using an artificial immunesystem [J].Journal of Network and Computer Applications.1996, 19(2):189-212[2]Tang Z, Yamaguchi T, and Tashima K, et al.. Multiple-valuedimmune network model and its simulations[C]. Proceedings ofthe 27th International Symposium on Multiple-valued logic,Autigonish, Canada, May 28-30, 1997: 233-238.[3]De Castro L N. An evolutionary immune network for dataclustering [C].VI Brazilian Symposium on Neural Networks(SBRN'00), Brazil, Jan. 22-25, 2000: 84-89.[4]Timmis J and Neal M. A resource limited artificial immunesystem for data analysis [J].Knowledge Based Systems.2001,14(3/4):121-130[5]Neal M. An artificial immune system for continuous analysisof time-varying data[C]. Proceedings of the 1st InternationalConference on Artificial Immune Systems Canterbury.Canterbury, UK, Sep. 9-11, 2002: 76-85.Nasaroui O, Gonzalez F, and Dasgupta D. The fuzzy artificialimmune system: Motivations, basic concepts, and applicationto clustering and web profiling[C]. Proceedings of the IEEEWorld Congress on Computational Intelligence, Honolulu,Hawaii, May 12-17, 2002: 711-716.[6]Xiong Hao and Sun Cai-xin. Artificial immune networkclassification algorithm for fault diagnosis of powertransformer [J].IEEE Transactions on Power Delivery.2007,22(2):930-935[7]Gu Dan-zhen, Ai Qian, and Chen Chen. The application ofartificial immune network in load classification[C].DRPT2008, Nanjing, China, Apirl 6-9 2008: 1394-1398.[8]L Chao and Hu Xiao-guang. Classification of mechanicalconditions for HVCBs based on artificial immune network[C].the 3rd IEEE on Industrial Electronics and Applications(ICIEA2008), Singapore, June 3-5, 2008: 2373-2377.[9]Li Zhong-hua, Zhang Yu-nong, and Tan Hong-zhou. Anefficient artificial immune network with elite-learning[C].IEEE Third International Conference on Naturalcomputation, Haikou, China, Aug. 24-27, 2007, 4: 213-217.[10]Li Liu and Xu Wen-bo. A cooperative artificial immunenetwork with particle swarm behavior for multimodalfunction optimization. Evolutionary Computation[C]. IEEEWorld Congress on Computational Intelligence. Washington,DC. July 22-24, 2008, 6: 1550-1555.[11]Fu Jian, Li Zhong-hua, and Tan Hong-zhou. A hybridartificial immune network with swarm learning [C].International Conference on Communications, Circuits andSystems, Kokura, Japan, July 11-13, 2007: 910-914.[12]钟燕飞, 张良培, 李平湘. 基于多值免疫网络的多光谱遥感影像分类[J]. 计算机学报, 2007, 30(12): 2181-2188.Zhong Yan-fei, Zhang Liang-pei and Li Ping-xiang.Classification of multi-Spectral remote sensing image basedon multiple-valued immune network [J]. Chinese Journal ofComputers, 2007, 30(12): 2181-2188.[13]Zhong Yang-fei, Zhang Liang-pei, and Gong Jian-ya, et al.. Asupervised artificial immune classifier for remote-sensingimagery [J].IEEE Transactions on Geosciences and RemoteSensing.2007, 45(12):3957-3966[14]Jerne N K. Towards a network theory of the immune system[J]. Ann. Immunol. (Inst. Pasteur) 1974, 125C: 373-389.[15]Leung K, Cheong F, and Cheong C. Generating compactclassifier systems using a simple artificial immune system [J].IEEE Transactions on System, Man and Cybernetics, Part B,2007, 10(5): 1344-1356.[16]Du Hai-feng, Jiao Li-cheng, and Wang Sun-an. Clonaloperator and antibody clone algorithms[C]. Proceedings ofthe First International Conference on Machine Learning andCybernetics, Beijing, 2002: 506-510.[17]Clausi D A and Yue B. Comparing concurrence probabilitiesand Markov fields for texture analysis of SAR sea ice image[J].IEEE Transactions on Geosciences and Remote Sensing.2004, 42(1):215-228[18]Mazidah puteh FTMSK, and Khairuddin Omar FTSM.Classifying heterogeneous data with artificial immunesystem[C]. International Symposium Information Technology,Shanghai, China, Dec. 21-22, 2008, 3: 1-5.

施引文献

资源附件(0)

访问统计