摘要： 针对样本间的不均衡性，提出一种基于成对约束的动态加权半监督模糊核聚类算法。在传统模糊聚类算法中加入半监督学习机制，通过Mercer核将原数据空间映射到特征空间，为特征空间中的每个向量分配一个动态权值，由此得到新的目标函数，并结合一种简单的核参数选择方法实现数据分类。理论分析和实验结果表明，与模糊核聚类算法及成对约束的竞争群算法相比，该算法具有更好的聚类效果。
关键词:
半监督聚类,
成对约束,
动态加权,
模糊聚类算法,
核参数
Abstract: In order to solve the imbalance between the data samples, this paper proposes a dynamic weighted semi-supervised fuzzy kernel clustering algorithm based on pairwise constraints, which incorporates both semi-supervised learning mechanism and the kernel function into traditional fuzzy clustering algorithm. Data in the original space are mapped to a high-dimensional feature space by Merce kernel functions, and an dynamic weight is assigned to each vector in the feature space. A modified objective function for fuzzy clustering is introduced in the feature space. A simple method is presented to determine the appropriate values for the kernel width. Theoretical analysis and experimental results testify that. the new algorithm has better clustering performance compared with other semi-supervised fuzzy clustering algorithms such as Pairwise Constrained Competitive Agglomeration(OCCA).
Key words:
semi-supervised clustering,
pairwise constraints,
dynamic weighted,
fuzzy clustering algorithm,
kernel parameter
中图分类号:
王亮, 王士同. 基于成对约束的动态加权半监督模糊核聚类[J]. 计算机工程, 2012, 38(01): 148-150.
WANG Liang, WANG Shi-Tong. Dynamic Weighted Semi-supervised Fuzzy Kernel Clustering Based on Pairwise Constraints[J]. Computer Engineering, 2012, 38(01): 148-150.