基于局部保留投影的稀疏中智聚类算法
Sparse Neutrosophic Clustering Algorithm Based on Local Preserving Projection
张丹 1马盈仓 1杨小飞 1邢志伟1
作者信息
- 1. 西安工程大学理学院 西安 710600
- 折叠
摘要
聚类算法是机器学习领域重要的研究课题之一,传统的中智聚类算法(例如FC-PFS算法)未考虑局部空间结构,且距离的计算受到冗余特征影响,不能有效处理高维数据集.为此,提出一种新的基于局部保留投影的稀疏中智聚类算法(LPSNCM)及其优化方法.一方面LPSNCM算法通过局部保留投影方法生成具有局部结构信息的正交投影空间,另一方面通过特征提取方法可以减少特征数量以获得更有效的特征,从而增强了FC-PFS算法处理高维数据的能力.LPSNCM算法也可以被看作是谱聚类两个独立阶段的统一模型.在一些基准数据集上的实验结果表明,与FC-PFS和某些最新方法相比,证明了LPSNCM的有效性.
Abstract
Clustering algorithm is one of the important research topics in the field of machine learning.Traditional neutrosoph-ic clustering algorithms(such as FC-PFS algorithm)do not consider the local spatial structure,and the calculation of distance is af-fected by redundant features and cannot effectively process high-dimensional data set.A new sparse neutrosophic clustering algo-rithm(LPSNCM)based on local preserved projection and its optimization method are proposed in this paper.On the one hand,an orthogonal projection space with local structure information is generated by the local preserved projection method in the LPSNCM al-gorithm,on the other hand,the feature extraction method can reduce the number of features to obtain more effective features,thus enhancing the capability of FC-PFS algorithm to process high-dimensional data.The LPSNCM algorithm can also be regarded as a unified model of the two independent stages of spectral clustering.Experimental results on some benchmark datasets demonstrate the effectiveness of LPSNCM compared with FC-PFS and some of the latest methods.
关键词
中智集/局部信息保留/基于投影的空间转化Key words
neutrosophic set/locality preserving projections/projection-based spatial transformation引用本文复制引用
基金项目
国家自然科学基金项目(61976130)
陕西省重点研发计划项目(2018KW-021)
陕西省自然科学基金项目(2020JQ-923)
出版年
2024
国家科技期刊平台
NSTL
万方数据