Scholarlink

<p>With the world heading towards online dating, frauds using dating websites are also increasing and have led to a huge surge in the creation of fake profiles, posing significant challenges to user trust and security. In this paper, we propose a novel approach to combating web-based fake profiles that incorporates an optimized version of the Light Gradient Boosting Machine (LightGBM) algorithm and the Synthetic Minority Over-sampling Technique (SMOTE) into the Obtain, Scrub, Explore, Model, and Interpret (OSEMN) framework. Enriched feature engineering was performed to obtain hybrid feature sets that include syntactic, semantic, linguistic, and user-based characteristics, and ensemble learning techniques were employed to significantly improve classification accuracy. We demonstrate the utility of our approach by correctly recognizing fake profiles and attaining an accuracy of 98.6% in substantially competent execution time, utilizing an optimized LightGBM algorithm with comprehensive hyperparameter settings. Furthermore, we conducted a comparative analysis with the MIB Twitter dataset, yielding promising results, thus demonstrating the flexibility and effectiveness of our method across various online platforms. The suggested framework joins hybrid feature engineering and hyperparameter optimization to improve the performance of the detection of fake profiles.</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

Combating web-based fake profiles using optimized LightGBM: an ensemble learning approach using OSEMN framework

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月21日

查看全文
  • Ahmad Kamal,
  • Lavina Pahuja,
  • Arpit Kumar Sharma

摘要

With the world heading towards online dating, frauds using dating websites are also increasing and have led to a huge surge in the creation of fake profiles, posing significant challenges to user trust and security. In this paper, we propose a novel approach to combating web-based fake profiles that incorporates an optimized version of the Light Gradient Boosting Machine (LightGBM) algorithm and the Synthetic Minority Over-sampling Technique (SMOTE) into the Obtain, Scrub, Explore, Model, and Interpret (OSEMN) framework. Enriched feature engineering was performed to obtain hybrid feature sets that include syntactic, semantic, linguistic, and user-based characteristics, and ensemble learning techniques were employed to significantly improve classification accuracy. We demonstrate the utility of our approach by correctly recognizing fake profiles and attaining an accuracy of 98.6% in substantially competent execution time, utilizing an optimized LightGBM algorithm with comprehensive hyperparameter settings. Furthermore, we conducted a comparative analysis with the MIB Twitter dataset, yielding promising results, thus demonstrating the flexibility and effectiveness of our method across various online platforms. The suggested framework joins hybrid feature engineering and hyperparameter optimization to improve the performance of the detection of fake profiles.