聚类数的自动确定

Automatic Determination of Number of Clusters

对聚类分析中聚类数的确定问题进行了研究。在SOFM神经网络的基础上,从聚类准则出发,通过试验对聚类准则的曲线特征进行了详细的分析和论证,设计出一种结构自适应的聚类神经网络,该网络能自动确定最佳的聚类数,并提出了一种减少计算量的改进算法。

Ascertain a problem to gathering the kind number＇s in clustering analysis having been in progress studying. On SOFM neural networks basis, the criterion sets off from clustering, analyse and expound arid prove detailedly by testing that the characteristic has been in progress to clustering criterion curve, design out one kind of certainly fit in with clustering of structure neural networks, be a network＇s turn to be able to have ascertained the optimum clustering number voluntarily, and have suggested that one kind cuts down the algorithm calculating the amounts improvement.