A negative selection algorithm for classification and reduction of the noise effect
A negative selection algorithm for classification and reduction of the noise effect
复制标题
DOI：
10.1016/j.asoc.2008.05.003
复制
发表时间：
2009
期刊：
影响因子：
--
通讯作者：
K. Igawa;H. Ohashi
中科院分区：
文献类型：
--
作者：
K. Igawa;H. Ohashi
人工免疫系统（AIS）是一种受人体免疫系统原理和过程启发的智能算法。在过去的十年中，AIS的应用已经在各个领域得到了研究。在变化/异常检测的应用中，AIS的负选择算法得到了成功的应用。然而，负选择算法并不适用于多类分类问题，因为它们没有一种机制来最小化过拟合和过度搜索的危险。在本文中，我们提出了一种新的算法来克服这一缺点，并将负选择算法的应用领域扩展到多类分类。我们提出的算法被称为人工负选择分类器（ANSC）。研究了ANSC对噪声的容忍度，提出了一种减小噪声对ANSC影响的方法。并与AIS分类器人工免疫识别系统（Artificial Immune Recognition System， AIRS）的准确率和数据约简进行了比较。结果表明，该算法对分类问题和降低噪声影响是有效的。
Artificial Immune Systems (AIS) are a type of intelligent algorithm inspired by the principles and processes of the human immune system. In the last decade, applications of AIS have been studied in various fields. In the application of change/anomaly detection, negative selection algorithms of AIS have been successfully applied. However, negative selection algorithms are not appropriate for multi-class classification problems, because they do not have a mechanism to minimize the danger of overfitting and oversearching. In this paper, we propose a new algorithm to overcome this drawback and to extend the application area of negative selection algorithms to multi-class classification. The algorithm we propose is named Artificial Negative Selection Classifier (ANSC). We investigate the tolerance of ANSC against noise, and introduce a method to reduce the effect of noise into ANSC. The accuracy and data reduction are compared with those from the Artificial Immune Recognition System (AIRS), which is a well known and effective classifier of AIS. The results show that our algorithm is useful for classification problems and the reduction of the noise effect.