Scholarlink

In this paper, we propose GMM-PACO, a novel Ant Colony Optimization (ACO) for multi-objective feature selection. This algorithm explicitly models each feature using Gaussian Mixture Models (GMMs) to capture distributional characteristics. Our heuristic function integrates mutual information with the Wasserstein distance between feature distributions to evaluate feature relevance and redundancy. GMM-PACO simultaneously optimizes classification accuracy, feature subset size, and redundancy using Pareto dominance. Extensive evaluations on multiple UCI datasets with k-nearest neighbors (KNN) classifier demonstrate improvements over existing methods in both subset compactness and classification accuracy.

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

GMM-PACO: Gaussian Mixture Models and Pareto-Based Ant Colony Optimization for Multi-objective Feature Selection

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Anna Krysta,
  • Inès Alaya,
  • Tristan Cazenave

摘要

In this paper, we propose GMM-PACO, a novel Ant Colony Optimization (ACO) for multi-objective feature selection. This algorithm explicitly models each feature using Gaussian Mixture Models (GMMs) to capture distributional characteristics. Our heuristic function integrates mutual information with the Wasserstein distance between feature distributions to evaluate feature relevance and redundancy. GMM-PACO simultaneously optimizes classification accuracy, feature subset size, and redundancy using Pareto dominance. Extensive evaluations on multiple UCI datasets with k-nearest neighbors (KNN) classifier demonstrate improvements over existing methods in both subset compactness and classification accuracy.