摘要
模糊C均值聚类算法使用欧氏距离衡量,遇到潜在的类或簇背离超球面结构时表现不佳。利用免疫理论中的克隆选择、亲和力成熟和免疫网络理论来建构一种网络模型aiNet,将其用于数据聚类可以减少数据中的冗余,描述数据结构和聚类形状。通过实验比较了这两种方法的特点,结果表明,当潜在的类或簇背离凸集时,aiNet方法表现出良好的适应性。
When using an Euclidean distance measure, fuzzy c-means clustering did not work well if the underlying classes or clusters deviated strongly from hyperspherical structures. An artificial network structure (aiNet) stressing the clonal selection and affinity maturation and immune network theory was capable of reducing redundancy, describing data structure, including the shape of clusters. The characters of two methods were compared by experiments, and results show that aiNet appears excellent adaptability when the underlying classes or clusters from convex set.
出处
《计算机工程与设计》
CSCD
2004年第4期515-517,588,共4页
Computer Engineering and Design
基金
湖南省自然科学基金项目(03JJY3101)