摘要
在详细分析克隆选择算法的基础上,提出非监督克隆选择聚类算法。该算法是数据驱动的自适应调整其参数,它对数据进行分类的操作尽可能快,改善过早收敛的问题,提高数据聚类的速度。通过使用一些人工和现实生活中的数据集,比较非监督克隆选择聚类算法与著名的K-means算法之间的性能优劣。实验结果表明,该算法不仅解决K-means算法需事先确定的类数K和在次佳值卡住的缺点,而且在功能上比传统的K-means分类算法具有较高的分类精度和更高的可靠性。
Based on the detailed analysis of clonal selection algorithm, proposes unsupervised clone selection clustering algorithm. Which is adaptive data driven by adjusting its parameters, it carries on the classification of data operations as soon as possible, improves the premature con- vergence problem, improves the speed of data clustering. By using several artificial and real-life data sets, comparing the performance between unsupervised clonal selection clustering algorithm K-means algorithm. The experimental results show that, this algorithm solves the K-means algorithm needs several classes of K determined in advance, and the second best value stuck faults, the classification accu- racy, and it is much better than traditional K-means classification algorithm in function and with higher reliability.
基金
广西科技大学鹿山学院科学基金项目(No.2013LSZK05)
关键词
人工免疫系统
克隆选择算法
聚类
多目标优化
K-MEANS算法
Artificial Immune Systems
Clonal Selection Algorithms
Clustering
Multi-Objective Optimization
K-means Algorithm