Scholarlink

Abstract <p>In this article, we propose a new approach of the K2 algorithm toenhance the process of selecting candidate parents in Bayesiannetworks by incorporating dependency criteria based on ridgeregression. Traditional K2 algorithms primarily rely on scoringfunctions to evaluate parent sets, which may not effectivelycapture all dependencies in the presence of multicollinearity. Ourapproach leverages ridge regression to address this limitation bypenalizing overly complex models, thereby providing a more robustmechanism for parent selection. Through real-world dataapplications, we demonstrate that our generalized algorithmsignificantly improves the accuracy and reliability of parentselection in Bayesian networks. The findings suggest that thisintegration of ridge regression with the K2 algorithm offers apromising avenue for advancing Bayesian network structurelearning, particularly in complex data.</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

A Generalized K2 Algorithm for Learning Bayesian Network Structures Using Ridge Regression

  • 期刊论文

  • 发表日期：

    2025年4月1日

查看全文
  • Mehryar Fallahnejad,
  • Vahid Rezaeitabar,
  • Mohammad Kazemi

摘要

Abstract

In this article, we propose a new approach of the K2 algorithm toenhance the process of selecting candidate parents in Bayesiannetworks by incorporating dependency criteria based on ridgeregression. Traditional K2 algorithms primarily rely on scoringfunctions to evaluate parent sets, which may not effectivelycapture all dependencies in the presence of multicollinearity. Ourapproach leverages ridge regression to address this limitation bypenalizing overly complex models, thereby providing a more robustmechanism for parent selection. Through real-world dataapplications, we demonstrate that our generalized algorithmsignificantly improves the accuracy and reliability of parentselection in Bayesian networks. The findings suggest that thisintegration of ridge regression with the K2 algorithm offers apromising avenue for advancing Bayesian network structurelearning, particularly in complex data.