基于阴影集的三支高斯混合聚类算法

Three-way Gaussian mixture clustering algorithm based on shadow set

针对三支高斯混合聚类算法(three-way Gaussian mixture model,T-GMM)的阈值通常为人为设定,增加算法的不确定性的问题,本文中将阴影集思想融入三支高斯混合模型,提出一种基于阴影集的三支高斯混合聚类算法(three-way Gaussian mixture model clustering based on shadow sets,ST-GMM);ST-GMM算法先构造一个关于阈值的目标函数,再通过优化算法选取最优阈值。基于10个不同类型的UCI数据集的实验结果表明:ST-GMM算法不仅继承了T-GMM算法的特点,同时有效地降低了人为设定阈值的误差,聚类细节的刻画也更加准确。针对评价指标的测试进一步验证了ST-GMM算法具有良好的聚类性能。

In order to solve the problem that three-way Gaussian mixture clustering algorithm(T-GMM)was usually set artificially,which increased the uncertainty of the algorithm,the shadow set was integrated into the three-way Gaussian mixture clustering algorithm,and a three-way Gaussian mixture clustering algorithm based on the shadow set ST-GMM was proposed.The new algorithm constructed an objective function about the threshold,and then selected the optimal threshold through the optimization algorithm.The experimental results on 10 different types of UCI data sets show that the proposed algorithm not only inherits the characteristics of T-GMM algorithm,but also effectively reduces the error caused by artificial threshold setting,and the description of clustering details is more accurate.The evaluation index tests further verify that the proposed algorithm has good clustering performance.