The effect of the uncertainties needs to be taken full advantage during uncertain data clustering.An uncertain data clustering algorithm based on fast Gaussian transform was proposed,to solve the problems about the impact on the accuracy of clustering results and the clustering efficiency caused by the uncertainties,during the construction of uncertain data models and the distance measurement,which existed in the current researches.First,the data model according to the characteristic of the uncertainty distribution was constructed,without the premise of assuming the data distribution.And the similarity between uncertain data objects was measured by combining the two important features of uncertain objects,attribute features and the probability density function representing the characteristic of uncertainty distribution.And then the uncertain data clustering algorithm was proposed.Finally,the experiment results on UCI and real datasets indicate the better efficiency and accuracy of proposed algorithm.%数据中不确定性的存在使对其聚类分析时要充分考虑不确定性的影响.针对现有不确定数据聚类算法中构建不确定数据模型以及距离度量时存在的影响结果准确性与聚类性能等问题,提出一种基于快速高斯变换的不确定数据聚类算法.首先在不假设数据分布的前提下,构建符合不确定性分布特征的数据模型;然后结合不确定对象的2个重要特征:属性特征与表示不确定数据分布特征的概率密度函数,度量不确定数据对象间的相似性;并以此为基础提出不确定数据聚类算法;最后在UCI以及真实数据集上的实验结果表明,所提算法在运行效率和聚类准确性方面均能取得较好效果.
展开▼