Scholarlink

E-commerce holds a pivotal position as the foundation of economic development, providing businesses with a variety of benefits. However, the increasing prevalence of fake news and misinformation has become a pressing issue, leading to negative consequences. This not only diminishes consumer trust but also threatens the integrity of the e-commerce landscape. The study utilizes the PhoBERT and LSTM deep learning model to detect fake news using five datasets: data1 (Vietnamese Twitter data), data2 (Vietnamese news data), data3 (English Twitter data), data4 (English social media data), and data5 (Vietnamese e-commerce data). Data2 was excluded due to overfitting, leaving four datasets for analysis. The experimental results demonstrate the effectiveness of the proposed models, achieving accuracy rates of 91% (data1), 85% (data3), 88% (data4), and 95% (data5), with corresponding AUC values of 87%, 75%, 79%, and 90%. Among them, there are three Vietnamese language datasets and two English language datasets. Specifically, the model performed exceptionally well on the Vietnamese e-commerce dataset (Data5), achieving a 95% accuracy, 90% AUC and an F1-score of 97% for real news, and 87% for fake news. This paper helps address the common issue of fake news, safeguarding the reputation and credibility of the e-commerce industry and promoting a safe online environment for consumers in Vietnam.

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 Deep Learning Approaches to Fake News Detection in E-commerce Platforms

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Thuy Kieu Phan Thi,
  • Quoc Hung Nguyen,
  • Hong Gam Le Thi,
  • Hoang Son Le

摘要

E-commerce holds a pivotal position as the foundation of economic development, providing businesses with a variety of benefits. However, the increasing prevalence of fake news and misinformation has become a pressing issue, leading to negative consequences. This not only diminishes consumer trust but also threatens the integrity of the e-commerce landscape. The study utilizes the PhoBERT and LSTM deep learning model to detect fake news using five datasets: data1 (Vietnamese Twitter data), data2 (Vietnamese news data), data3 (English Twitter data), data4 (English social media data), and data5 (Vietnamese e-commerce data). Data2 was excluded due to overfitting, leaving four datasets for analysis. The experimental results demonstrate the effectiveness of the proposed models, achieving accuracy rates of 91% (data1), 85% (data3), 88% (data4), and 95% (data5), with corresponding AUC values of 87%, 75%, 79%, and 90%. Among them, there are three Vietnamese language datasets and two English language datasets. Specifically, the model performed exceptionally well on the Vietnamese e-commerce dataset (Data5), achieving a 95% accuracy, 90% AUC and an F1-score of 97% for real news, and 87% for fake news. This paper helps address the common issue of fake news, safeguarding the reputation and credibility of the e-commerce industry and promoting a safe online environment for consumers in Vietnam.