Scholarlink

In the past few years’ violent conflicts between communities on grounds of political ideology has become quite evident. Social media algorithms frequently personalize material based on user choices, resulting in echo chambers in which users are sub consciously made to confirm their preconceptions. However, limited efforts have been made to understand and make a filter to curb this misinformation. This work seeks to evaluate potency of five modern algorithms, i.e. Support Vector Machine algorithm (SVM), Naive Bayes Classifier (NBC), Random Forest Classifier (RFC), Decision Tree Classifier (DTC) and Long Short-Term Memory algorithm (LSTM). The paper assesses the effectiveness of various algorithms i.e. on 4 different datasets from authentic sources which include news title, news description, and its label of being false/true. Every model is cleaned before evaluation for best results. Six criteria were used to assess the algorithms (a) precision (b) recall (c) accuracy (d) F1 score (e) Confusion matrix and (f) dataset richness. This detection prevents privacy breaches, ensuring personal data security by protecting from false narratives.

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

Dealing with Fake News: Machine-Learning Approaches to News Classification

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Saksham Singh,
  • Ankita Gupta

摘要

In the past few years’ violent conflicts between communities on grounds of political ideology has become quite evident. Social media algorithms frequently personalize material based on user choices, resulting in echo chambers in which users are sub consciously made to confirm their preconceptions. However, limited efforts have been made to understand and make a filter to curb this misinformation. This work seeks to evaluate potency of five modern algorithms, i.e. Support Vector Machine algorithm (SVM), Naive Bayes Classifier (NBC), Random Forest Classifier (RFC), Decision Tree Classifier (DTC) and Long Short-Term Memory algorithm (LSTM). The paper assesses the effectiveness of various algorithms i.e. on 4 different datasets from authentic sources which include news title, news description, and its label of being false/true. Every model is cleaned before evaluation for best results. Six criteria were used to assess the algorithms (a) precision (b) recall (c) accuracy (d) F1 score (e) Confusion matrix and (f) dataset richness. This detection prevents privacy breaches, ensuring personal data security by protecting from false narratives.