| 摘要: |
| 针对数据简化中的实例选择问题,基于抗体克隆选择学说提出了一种免疫克隆数据简化算法.利用马尔可夫理论证明了该算法能以概率1 收敛.通过对7 个具有代表性的标准UCI 数据集的简化实验证明了该算法的有效性.通过实验分析了权值参数λ的取值变化对算法性能的影响,确定了其最佳取值区间.针对海量数据集简化时算法收敛较慢的问题,引入分层编码策略.通过对7 个大规模及海量数据集的简化实验表明了在进化代数不变的情况下,新的编码方式能够极大地提高算法的收敛速度,得到更为理想的结果.通过对Letter 和DNA两个数据集的实验给出了分层编码中层数t的最佳取值区间. |
| 关键词: 克隆选择 数据简化 实例选择 人工免疫系统 进化计算 |
| DOI： |
| 分类号: |
| 基金项目:Supported by the National Natural Science Foundation of China under Grant Nos.60703107, 60703108 (国家自然科学基金); the National High-Tech Research and Development Plan of China under Grant No.2009AA12Z210 (国家高技术研究发展计划(863)); the Program for New Century Excellent Talents in University of China (新世纪优秀人才支持计划); the National Basic Research Program of China under Grant No.2006CB705700 (国家重点基础研究发展计划(973)) |
|
| Data Reduction Based on Artificial Immune System |
|
GONG Mao-Guo,HAO Lin,JIAO Li-Cheng,WANG Xiao-Hua,SUN Yi-Fei
|
| Abstract: |
| Based on the antibody clonal selection theory, an immune clonal data reduction algorithm is proposed
for instance selection problems of data reduction. The theory of Markov chain proves that the new algorithm is convergent with probability 1. The experimental studies on seven standard data sets of UCI repository show that the algorithm proposed in this paper is effective. The best domain of the weight parameter λ is determined by analyzing its effect on algorithm’s performance. Furthermore, an encoding method based on the stratified strategy is introduced to accelerate the convergence speed when solving large scale data reduction problems. The experimental studies based on seven large scale data sets show that the improved method is superior to the primary one. Finally,the best domain of the number of stratums t is determined by analyzing its effect on algorithm’s performance based on the data sets Letter and DNA. |
| Key words: clonal selection data reduction instance selection artificial immune system evolutionary
computation |
