摘要
为了克服传统K-Means算法k值不能确定问题和不具备变量自动选择能力,将预测强度和变量自动加权K-Means算法相结合,提出基于预测强度的变量自动加权K-Means算法。预测强度表示聚类模型对未知数据的预测能力,预测能力越强,则聚类结果越佳,主要用于k值的确定;变量自动加权K-Means算法具有在聚类过程中自动调整变量权重的能力,对于噪声变量和冗余变量削弱其对距离的贡献,使聚类结果反映最真实的聚类结构。实验表明,算法具有较强的分类能力和预测能力。
In order to overcome the problems of unknown k value before clustering and less properties automatically choosing ability of conventional K-Means algorithm,automatically variable weighting K-Means algorithm based on forecasting Intensity is proposed. Forecasting Intensity indicates the forecasting capability of training mode to those extra unknown data sample. Better the forecasting is,better the clustering results are. It was mainly used to get the k value of clustering algorithm. Automatically variable weighting K-Means algorithm could adjust the weights of each properties. This can reduce the distance which noise properties and redundant properties contributed in the processing of clustering and reflect the real cluster structure. This article combines forecasting intensity, automatically variable weighting K-Means algorithm and empirical results show its classification ability and forecasting ability.
出处
《四川理工学院学报（自然科学版）》
CAS
2016年第2期25-29,共5页
Journal of Sichuan University of Science & Engineering(Natural Science Edition)
关键词
K-MEANS
预测强度
变量自动加权
K-Means
forecasting intensity
automatically variable weighting
