【作者】 高翠芳； 吴小俊；

【Author】 Cui-Fang GAO,Xiao-jun WU (School of Information Engineering,Jiangnan University,Wuxi 214122,China )

【机构】 江南大学信息工程学院；

【摘要】 聚类数目是影响聚类效果的关键参数,通常需要人工确定,对于较难获得这一先验知识的数据集,聚类分析会因此受到限制。针对这一问题,文章提出了一种自动确定最佳聚类数目的方法,该方法利用体现"类内紧凑类间离散"思想的优化聚类算法来执行主要计算,结合目标函数二阶差分的判定准则,通过聚类算法的自学习来确定最佳聚类数。实验结果显示,该方法能在不同数据集上自动得到正确的聚类数目。更多还原

【Abstract】 The number of cluster,which is generally determined manually,is a critical parameter affecting the effectiveness of clustering.So the clustering techniques are not applicable in some datasets that are difficult to get such apriori information.To overcome this problem,an automatic technique to determine the optimal cluster number was proposed in this paper.The significant calculation is implemented by an optimization algorithm which utilizes the reformulated objective function that reflects the idea of compactness of intra-cluster and separation of inter-cluster,and then the optimal cluster number is determined by using the maximum criteria of second order difference of objective function.The simulated tests show that the new method can automatically and properly get the cluster number for different datasets.更多还原