Scholarlink

<p>Blogging has emerged as a prominent platform for internet users to express their opinions. However, this open dialogue is susceptible to spam comments that disseminate irrelevant, abusive, and unsolicited content, potentially spreading false information. To address this challenge, an Efficient Spam comment detection (E-SCD) model is proposed in three key steps advanced preprocessing, hybrid deep learning (DL), and ensemble classification to enhance detection accuracy. The approach integrates a Convolutional Neural Network (CNN) for spatial and semantic feature extraction and a Bidirectional Long Short-Term Memory (BiLSTM) network for capturing temporal dependencies. These features are then classified using Extreme Gradient Boosting (XGBoost), ensuring robust performance. Experiments on a real-world blog dataset demonstrate the model’s superior accuracy 93.8%, precision 93.8%, recall 93.8%, F1-score 93.7%, and AUC 93.05%. To further validate generalizability, the model was evaluated on two additional publicly available datasets, HuggingFace Spam Detection and YouTube Comments Spam. Across all three datasets, the proposed model consistently achieved the highest performance, outperforming baseline Machine Learning (ML) and DL models. The results highlight the potential of hybrid ensemble methods for real-time, scalable spam detection in dynamic comment environments. The model’s superior efficiency surpasses current standards, providing a tangible advantage in preventing spam activities in online conversations.</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

Towards efficient deep learning assisted ensemble approach for spam comments detection in blogs data

  • 期刊论文

  • 发表日期：

    2026年5月30日

查看全文
  • Amreen Batool,
  • Hasnain Hyder,
  • Yung-Cheol Byun

摘要

Blogging has emerged as a prominent platform for internet users to express their opinions. However, this open dialogue is susceptible to spam comments that disseminate irrelevant, abusive, and unsolicited content, potentially spreading false information. To address this challenge, an Efficient Spam comment detection (E-SCD) model is proposed in three key steps advanced preprocessing, hybrid deep learning (DL), and ensemble classification to enhance detection accuracy. The approach integrates a Convolutional Neural Network (CNN) for spatial and semantic feature extraction and a Bidirectional Long Short-Term Memory (BiLSTM) network for capturing temporal dependencies. These features are then classified using Extreme Gradient Boosting (XGBoost), ensuring robust performance. Experiments on a real-world blog dataset demonstrate the model’s superior accuracy 93.8%, precision 93.8%, recall 93.8%, F1-score 93.7%, and AUC 93.05%. To further validate generalizability, the model was evaluated on two additional publicly available datasets, HuggingFace Spam Detection and YouTube Comments Spam. Across all three datasets, the proposed model consistently achieved the highest performance, outperforming baseline Machine Learning (ML) and DL models. The results highlight the potential of hybrid ensemble methods for real-time, scalable spam detection in dynamic comment environments. The model’s superior efficiency surpasses current standards, providing a tangible advantage in preventing spam activities in online conversations.