Scholarlink

<p>In the fields of machine learning and data mining, feature selection (FS) plays a vital role in identifying an appropriate subset of features from a set of high-dimensional, potentially irrelevant, and redundant features. A heterogeneous information system (HIS) is characterized by the integration of three distinct types of feature information values: scaled types, ordered types, and normal types. This study initially establishes a model for a <i>k</i>-nearest neighborhood rough set by defining the distance metrics for different types of data and subsequently employs this model to design rough information granulation. The research then introduces opposition-based learning to particle swarm optimization (PSO) algorithm to address FS tasks. Following this, a FS algorithm in HIS, referred to as FSRSPSO, is developed by utilizing a combination of opposition-based PSO and rough information granulation. To evaluate the efficacy and reliability of the FSRSPSO, clustering methods are applied to both the raw heterogeneous data and the feature subsets identified by FSRSPSO. The experimental findings demonstrate that FSRSPSO represents a dependable and efficient feature selection algorithm in HIS.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Opposition-based particle swarm optimization algorithm integrating neighborhood rough set for feature selection in heterogeneous information system

  • 期刊论文

  • 发表日期：

    2025年5月18日

查看全文
  • Jie Zhang,
  • Zhijun Zhao

摘要

In the fields of machine learning and data mining, feature selection (FS) plays a vital role in identifying an appropriate subset of features from a set of high-dimensional, potentially irrelevant, and redundant features. A heterogeneous information system (HIS) is characterized by the integration of three distinct types of feature information values: scaled types, ordered types, and normal types. This study initially establishes a model for a k-nearest neighborhood rough set by defining the distance metrics for different types of data and subsequently employs this model to design rough information granulation. The research then introduces opposition-based learning to particle swarm optimization (PSO) algorithm to address FS tasks. Following this, a FS algorithm in HIS, referred to as FSRSPSO, is developed by utilizing a combination of opposition-based PSO and rough information granulation. To evaluate the efficacy and reliability of the FSRSPSO, clustering methods are applied to both the raw heterogeneous data and the feature subsets identified by FSRSPSO. The experimental findings demonstrate that FSRSPSO represents a dependable and efficient feature selection algorithm in HIS.