Scholarlink

<p>Sentiment analysis or opinion mining is crucial to extracting relevant information from text documents based on various sources. Numerous techniques, including lexicon-based, machine learning algorithms, and rule-based, are used for sentiment analysis. However, lexicon-based sentiment classification faces low accuracies due to domain-oriented competitive dictionary deficiencies, while machine learning-based sentiment faces accuracy constraints due to feature ambiguity. To address this issue, a stacking-based ensemble learning technique is employed to predict sentiments from Twitter data. The Twitter sentiment analysis dataset is first used to gather raw data, which is pre-processed to eliminate any superfluous information. After pre-processing, the features of the text are extracted using the Bidirectional Encoder Representations from Transformers (BERT) technique, and its parameters are tuned using the Fire Hawk Optimizer (FHO) optimization algorithm. The Pearson correlation coefficient is used to select the features. The selected features are fed into deep-belief neural networks and modular neural networks. The output of both classifiers is given to a Meta classifier called Jordan Neural Network (JNN) to effectively predict user sentiment based on Twitter’s big data. According to simulated research, the proposed approach achieves 97.49% accuracy, 2.51% error, and 95.74% precision. Consequently, the proposed approach outperforms other current methods. Therefore, the model’s design anticipated sentiments of Twitter big data in an effective manner.</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

Stacking based ensemble learning and BERT with FHO approach for analysing sentiment in Twitter big data

  • 期刊论文

  • 发表日期：

    2025年10月15日

查看全文
  • Bondili Naga Sai Bhavya Charitha,
  • Ramanchi Radhika

摘要

Sentiment analysis or opinion mining is crucial to extracting relevant information from text documents based on various sources. Numerous techniques, including lexicon-based, machine learning algorithms, and rule-based, are used for sentiment analysis. However, lexicon-based sentiment classification faces low accuracies due to domain-oriented competitive dictionary deficiencies, while machine learning-based sentiment faces accuracy constraints due to feature ambiguity. To address this issue, a stacking-based ensemble learning technique is employed to predict sentiments from Twitter data. The Twitter sentiment analysis dataset is first used to gather raw data, which is pre-processed to eliminate any superfluous information. After pre-processing, the features of the text are extracted using the Bidirectional Encoder Representations from Transformers (BERT) technique, and its parameters are tuned using the Fire Hawk Optimizer (FHO) optimization algorithm. The Pearson correlation coefficient is used to select the features. The selected features are fed into deep-belief neural networks and modular neural networks. The output of both classifiers is given to a Meta classifier called Jordan Neural Network (JNN) to effectively predict user sentiment based on Twitter’s big data. According to simulated research, the proposed approach achieves 97.49% accuracy, 2.51% error, and 95.74% precision. Consequently, the proposed approach outperforms other current methods. Therefore, the model’s design anticipated sentiments of Twitter big data in an effective manner.