摘要
由于非空间属性维数较高,空间聚类算法在处理非空间属性约束时难点首先在于如何为这些非空间属性设定参数,然后是哪些非空间属性在聚类中将起主要作用,并真正影响聚类的结果。对这些问题进行了讨论,并提出使用神经网络中自组织映射的方法来首先选择哪些非空间属性将被优先考虑,使用自组织特征映射(SOFM)方法对非空间属性聚类,最后把非空间属性和空间属性聚类进行合并得到最终的聚类结果的方法。
Because of high dimension characteristic of non-spatial attributes, the difficulties in operation are how to set parameters for these attributes. When using general spatial clustering algorithm, the difficulties are how to judge which dimensions play main role and affect cluster result. Based on the research of those problems, a method for analyzing nonspatial attributes was proposed. First, Self-Organizing Feature Map （SOFM） was adopted to choose some dimensions, and used to cluster the dense non-spatial attributes on these dimensions. Then the cluster of non-spatial attributes and that of spatial attributes were merged.
出处
《计算机应用》
CSCD
北大核心
2006年第11期2667-2669,2673,共4页
journal of Computer Applications
关键词
聚类算法
高维
神经网络
自组织特征映射
约束
clustering algorithm
high-dimension
neural network
Self-Organizing Feature Map（SOFM）
constraint
