Scholarlink

<p>Addressing feature interactions is essential in binary classification issues. This paper explores methods for capturing complete weighted feature interactions under two distinct assumptions: independent and dependent importance. Building upon the normalized interaction-weighted HM (NIWHM) and Maclaurin mean (NIWMM) operators, we introduce the normalized interaction-weighted Heronian–Maclaurin mean (NIWHMM) and the normalized weighted Heronian–Maclaurin mean (NWHMM) operators. These operators are further generalized into the weighted Heronian–Maclaurin mean (WHMM) operators. Subsequently, we propose a WHMM-based classifier to capture complete weighted feature interactions with independent importance. Furthermore, drawing inspiration from the factorization machine (FM) and its associated classifiers, we propose an FM-WHMM-based classifier to capture complete weighted feature interactions with dependent importance. Extensive experiments on synthetic and real-world datasets validate the effectiveness of the proposed classifiers. Given their demonstrated ability to capture complete weighted feature interactions, the proposed methods are considered to have promising application potential in fields requiring feature interactions, such as recommendation systems and sentiment analysis.</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

Weighted Heronian–Maclaurin mean-based classifiers combined with FM and their application in binary classification

  • 期刊论文

  • 发表日期：

    2025年8月20日

查看全文
  • Jia Han Wang,
  • Zhen Ming Ma,
  • Wen Hui Zhang,
  • Wei Yang

摘要

Addressing feature interactions is essential in binary classification issues. This paper explores methods for capturing complete weighted feature interactions under two distinct assumptions: independent and dependent importance. Building upon the normalized interaction-weighted HM (NIWHM) and Maclaurin mean (NIWMM) operators, we introduce the normalized interaction-weighted Heronian–Maclaurin mean (NIWHMM) and the normalized weighted Heronian–Maclaurin mean (NWHMM) operators. These operators are further generalized into the weighted Heronian–Maclaurin mean (WHMM) operators. Subsequently, we propose a WHMM-based classifier to capture complete weighted feature interactions with independent importance. Furthermore, drawing inspiration from the factorization machine (FM) and its associated classifiers, we propose an FM-WHMM-based classifier to capture complete weighted feature interactions with dependent importance. Extensive experiments on synthetic and real-world datasets validate the effectiveness of the proposed classifiers. Given their demonstrated ability to capture complete weighted feature interactions, the proposed methods are considered to have promising application potential in fields requiring feature interactions, such as recommendation systems and sentiment analysis.