Scholarlink

<p>This paper introduces a new hybrid method for improving the accuracy of sentiment classification for a drug review dataset in the context of drug reviews. It is a dataset for user-generated reviews about prescription and over-the-counter medications, covering aspects like efficacy, side effects, and patient satisfaction ratings. In order to tackle issues such as class imbalance and the scarcity of labeled data, we propose an architecture consisting of conventional synonym-based data augmentation methodology followed by the application of generative adversarial networks (GANs). Synthetic reviews are generated with the GAN model, allowing for greater diversity in the underlying data, increasing the accuracy of sentiment classification. Logistic regression was used with hyperparameter optimization to evaluate the performance of the augmented and synthetic dataset for classification tasks. The performance evaluation results show that the proposed hybrid model outperforms the baseline with an accuracy of 86%, precision of 85%, and recall of 82% and F1-score of 83%. Such enhancements lead to improved practices in pharmacovigilance and a more comprehensive understanding of drug safety and effectiveness. The proposed approach has the potential to improve drug surveillance systems, clinical decision-making, and marketing strategies in the pharmaceutical sector.</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

Integrating Synonym-Based Data Augmentation and Generative Adversarial Networks for Enhancing Adverse Drug Event Prediction

  • 期刊论文

  • 发表日期：

    2026年5月13日

查看全文
  • Mani Butwall,
  • Rahul Kumar Vijay

摘要

This paper introduces a new hybrid method for improving the accuracy of sentiment classification for a drug review dataset in the context of drug reviews. It is a dataset for user-generated reviews about prescription and over-the-counter medications, covering aspects like efficacy, side effects, and patient satisfaction ratings. In order to tackle issues such as class imbalance and the scarcity of labeled data, we propose an architecture consisting of conventional synonym-based data augmentation methodology followed by the application of generative adversarial networks (GANs). Synthetic reviews are generated with the GAN model, allowing for greater diversity in the underlying data, increasing the accuracy of sentiment classification. Logistic regression was used with hyperparameter optimization to evaluate the performance of the augmented and synthetic dataset for classification tasks. The performance evaluation results show that the proposed hybrid model outperforms the baseline with an accuracy of 86%, precision of 85%, and recall of 82% and F1-score of 83%. Such enhancements lead to improved practices in pharmacovigilance and a more comprehensive understanding of drug safety and effectiveness. The proposed approach has the potential to improve drug surveillance systems, clinical decision-making, and marketing strategies in the pharmaceutical sector.