Scholarlink

Social networks require accurate community detection to analyze information flow and network behavior, but traditional methods often struggle with large-scale, complex structures. To address this, we propose DWECD, a deep learning-based algorithm that integrates an enhanced random walk model, the PPMI matrix, and a Stacked Denoising Autoencoder to learn robust node embeddings, followed by K-means clustering. Experiments on real-world datasets show that DWECD significantly outperforms traditional methods. Specifically, compared with Spectral Clustering, DWECD achieves about 12% higher modularity (Q), 140% higher NMI, and 56% higher F-score on the Football dataset, and about 20% higher modularity, 25-fold higher NMI, and 130% higher F-score on the YouTube dataset, offering an effective and scalable solution for large-scale community detection.

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

The Community Detection Algorithm Based on Deep Walk Embedding

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Cheng Xue,
  • Rong Fei,
  • Hailong Peng

摘要

Social networks require accurate community detection to analyze information flow and network behavior, but traditional methods often struggle with large-scale, complex structures. To address this, we propose DWECD, a deep learning-based algorithm that integrates an enhanced random walk model, the PPMI matrix, and a Stacked Denoising Autoencoder to learn robust node embeddings, followed by K-means clustering. Experiments on real-world datasets show that DWECD significantly outperforms traditional methods. Specifically, compared with Spectral Clustering, DWECD achieves about 12% higher modularity (Q), 140% higher NMI, and 56% higher F-score on the Football dataset, and about 20% higher modularity, 25-fold higher NMI, and 130% higher F-score on the YouTube dataset, offering an effective and scalable solution for large-scale community detection.