Scholarlink

Abstract <p>The growing complexity of energy consumption data in urban environments poses significant challenges for conventional clustering methods, particularly the K-means algorithm, due to its sensitivity to initialization and its tendency to converge toward local optima. This research addresses these limitations by proposing a hybrid clustering approach that combines Ant Colony Optimization (ACO) with statistical mode-based initialization. The methodology combines deterministic statistical initialization to stabilize the early stages of clustering with bio-inspired global exploration to refine ambiguous assignments. The proposed K‑means/ACO/mode model is evaluated on two datasets for clustering energy consumption patterns with varying scales and characteristics. Performance is evaluated using the silhouette index, entropy, and computational cost, including reproducibility analysis across multiple runs and with different numbers of clusters. The experimental results demonstrate that the hybrid approach consistently outperforms conventional K-means, even when initialized with mode and K-means combined with ACO alone. In particular, the proposed method achieves greater cluster cohesion, significantly reduces uncertainty, and improves stability at different values of K, while maintaining a reasonable computational cost. The proposed model provides a reliable, reproducible clustering framework with direct applicability to energy-efficiency analysis and decision-making in the context of smart cities.</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 Hybrid Approach based on Ant Colony Optimization to Improve the Performance of the K-Means Algorithm for Clustering Energy Efficiency Patterns

  • 期刊论文

  • 发表日期：

    2026年6月23日

查看全文
  • Sinuhe Ginés-Palestino,
  • Eduardo Roldán-Reyes,
  • Marcela Quiroz-Castellanos,
  • Consuelo Gines Palestino,
  • Paulo Nazareno Maia Sampaio

摘要

Abstract

The growing complexity of energy consumption data in urban environments poses significant challenges for conventional clustering methods, particularly the K-means algorithm, due to its sensitivity to initialization and its tendency to converge toward local optima. This research addresses these limitations by proposing a hybrid clustering approach that combines Ant Colony Optimization (ACO) with statistical mode-based initialization. The methodology combines deterministic statistical initialization to stabilize the early stages of clustering with bio-inspired global exploration to refine ambiguous assignments. The proposed K‑means/ACO/mode model is evaluated on two datasets for clustering energy consumption patterns with varying scales and characteristics. Performance is evaluated using the silhouette index, entropy, and computational cost, including reproducibility analysis across multiple runs and with different numbers of clusters. The experimental results demonstrate that the hybrid approach consistently outperforms conventional K-means, even when initialized with mode and K-means combined with ACO alone. In particular, the proposed method achieves greater cluster cohesion, significantly reduces uncertainty, and improves stability at different values of K, while maintaining a reasonable computational cost. The proposed model provides a reliable, reproducible clustering framework with direct applicability to energy-efficiency analysis and decision-making in the context of smart cities.