Scholarlink

This paper presents an enhanced approach to sentiment analysis on Twitter data, leveraging recent advancements in machine learning and deep learning. In addition to traditional classifiers, such as Logistic Regression, Naive Bayes, SVM, and Random Forest, we incorporate Transformer-based models like BERT to achieve improved performance. Furthermore, the work explores hybrid techniques combining machine learning models with lexicon-based methods and hyperparameter optimization. Ultimately, we incorporated an ensemble learning approach which provided the best performance by leveraging the strengths of multiple models. We tested the models using various metrics like F1-score and ROC-AUC. Our results show that ensemble models along with hybrid techniques perform much better than traditional methods at classifying tweets as positive, negative, or neutral.

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

Improving Twitter Sentiment Analysis with a Hybrid BERT and Graph Neural Network Ensemble

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Amar Taggu,
  • Nabam Teyi

摘要

This paper presents an enhanced approach to sentiment analysis on Twitter data, leveraging recent advancements in machine learning and deep learning. In addition to traditional classifiers, such as Logistic Regression, Naive Bayes, SVM, and Random Forest, we incorporate Transformer-based models like BERT to achieve improved performance. Furthermore, the work explores hybrid techniques combining machine learning models with lexicon-based methods and hyperparameter optimization. Ultimately, we incorporated an ensemble learning approach which provided the best performance by leveraging the strengths of multiple models. We tested the models using various metrics like F1-score and ROC-AUC. Our results show that ensemble models along with hybrid techniques perform much better than traditional methods at classifying tweets as positive, negative, or neutral.