【摘要】 传统的聚类算法如k-means算法需要一些先验知识来确定初始参数,初始参数的选择通常会对聚类结果生产很大的影响。提出一种新的基于模型的聚类算法,通过优化给定的数据和数学模型之间的适应性发现数据对模型的最好匹配。由于高斯混合模型可以看作是一种“软分配聚类”方法,该算法结合一种贪心的EM算法来学习高斯混合模型(GMM),由贪心EM算法实现高斯混合模型结构和参数的自动学习,而不需要先验知识。这种聚类算法可以克服k-means等算法的缺点,实验结果表明该算法具有更好的聚类效果。更多还原

【Abstract】 Traditional clustering algorithms such as k-means algorithm require some prior knowledge to determine initial parameters, but the selection of initial parameters usually influences the clustering result greatly. In this paper, a new clustering algorithm based on model is proposed, it finds the best fit of the given data to mathematical model by optimizing the fit between the data and model. GMM can be treated as a “soft assignment clustering” method, and this algorithm learns GMM by combining a greedy EM algorithm which has the ability to learn the GMM structure and parameters automatically without any requirement for prior knowledge. This clustering algorithm can overcome the disadvantage of k-means algorithm and the experiments have been implemented to evaluate the efficiency and performance of the algorithm.更多还原