基于空间结构的符号数据仿射传播算法
发布时间：2018-08-08 16:29
【摘要】：由于符号型数据缺乏清晰的空间结构,很难构造一种合理的相似性度量,从而使诸多数值型聚类算法难以推广至符号型数据聚类.基于此种情况,文中引入一种空间结构表示方法,把符号型数据转化为数值型数据,能够在保持原符号型数据的结构特征的基础上重新构造样本之间的相似度.基于此方法,将仿射传播(AP)聚类算法迁移至符号数据聚类中,提出基于空间结构的符号数据AP算法(SBAP).在UCI数据集中若干符号型数据集上的实验表明,SBAP可以使AP算法有效处理符号型数据聚类问题,并且可以提升算法性能.
[Abstract]:Due to the lack of clear spatial structure of symbolic data, it is difficult to construct a reasonable similarity measure, which makes it difficult to generalize many numerical clustering algorithms to symbolic data clustering. In this case, a spatial structure representation method is introduced in this paper, which converts symbolic data into numerical data, which can reconstruct the similarity between samples on the basis of preserving the structural characteristics of the original symbolic data. Based on this method, the affine propagating (AP) clustering algorithm is migrated to symbol data clustering, and a spatial structure-based symbol data AP algorithm (SBAP). Is proposed. Experiments on several symbolic data sets in UCI dataset show that the AP algorithm can effectively deal with the symbolic data clustering problem and improve the performance of the algorithm.
【作者单位】： 山西大学计算机与信息技术学院;山西大学计算智能与中文信息处理教育部重点实验室;
【基金】：国家自然科学基金项目(No.61432011,U1435212,61322211) 教育部新世纪优秀人才支持计划(No.NCET-12-1031) 高等学校博士学科点专项科研基金(博导类)(No.20121401110013) 山西省高等学校优秀青年学术带头人项目(No.20120301)资助~~
【分类号】：TP311.13
[Abstract]:Due to the lack of clear spatial structure of symbolic data, it is difficult to construct a reasonable similarity measure, which makes it difficult to generalize many numerical clustering algorithms to symbolic data clustering. In this case, a spatial structure representation method is introduced in this paper, which converts symbolic data into numerical data, which can reconstruct the similarity between samples on the basis of preserving the structural characteristics of the original symbolic data. Based on this method, the affine propagating (AP) clustering algorithm is migrated to symbol data clustering, and a spatial structure-based symbol data AP algorithm (SBAP). Is proposed. Experiments on several symbolic data sets in UCI dataset show that the AP algorithm can effectively deal with the symbolic data clustering problem and improve the performance of the algorithm.
【作者单位】： 山西大学计算机与信息技术学院;山西大学计算智能与中文信息处理教育部重点实验室;
【基金】：国家自然科学基金项目(No.61432011,U1435212,61322211) 教育部新世纪优秀人才支持计划(No.NCET-12-1031) 高等学校博士学科点专项科研基金(博导类)(No.20121401110013) 山西省高等学校优秀青年学术带头人项目(No.20120301)资助~~
【分类号】：TP311.13
【相似文献】
相关期刊论文 前10条
1 徐结绿,徐汉良,吕述望;仿射全向置换的构造和计数[J];通信技术;2003年05期
2 龚石钰;;两平面场仿射及其在工程上的应用[J];成都科技大学学报;1989年06期
3 李天宝,陈文波,石世宏;仿射图形的计算机作图方法的研究[J];南华大学学报(理工版);2003年01期
4 刘黎,董培蓓;平行线束法的仿射研究[J];工程图学学报;2004年04期
5 张青,李永慈,唐守正;基于仿射重构的树高测量[J];计算机工程与应用;2005年31期
6 张桂梅;任伟;储s，
本文编号：2172363
本文链接：https://www.wllwen.com/kejilunwen/ruanjiangongchenglunwen/2172363.html
