一种改进的基于特征赋权的K均值聚类算法

An Improved K-Means Clustering Algorithm Based on Feature Weighting

聚类分析是数据挖掘及机器学习领域内的重点问题之一。近年来，为了提高聚类质量，借鉴和引入了分类领域特征选择及特征赋权思想，提出了一些基于特征赋权的聚类算法。在这些研究基础上，本文提出了一种基于密度的初始中心点选择算法，并借鉴文[1]所提出的特征赋权方法，给出了一种改进的基于特征赋权的K均值算法。实验表明该算法能较为稳定地得到较高质量的聚类结果。

Clustering analysis is one of the important problems in the data mining and machine learning areas. Recently, feature selection and feature weighting methods are introduced to clustering algorithms for improving the clustering quality. Inspired by the research, an improved k-means clustering based on feature weighting is proposed, which proposes a density-based initial centers search algorithm. The experiments show that the proposed algorithm can result in high quality clustering steadily.