Scholarlink

When training the automatic classification and grading model for metadata from power grid marketing system, challenges may arise due to inconsistent data specifications and imbalanced categories. Given that data balance significantly impacts the effectiveness of model training, it is typically necessary to clean and annotate manually. This paper proposes a deep learning-based data augmentation algorithm combined with generative adversarial network. By leveraging a small set of clean data, this approach generates additional samples to augment the dataset, thereby achieving the objective of enhancing data diversity.

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

An Enhancement Algorithm for Category-Imbalanced Data in Power Grid Marketing System

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Xuhua Ai,
  • Yiting Huang,
  • Yuan Yin,
  • Yun Dong,
  • Qi Meng,
  • Xixiang Zhang

摘要

When training the automatic classification and grading model for metadata from power grid marketing system, challenges may arise due to inconsistent data specifications and imbalanced categories. Given that data balance significantly impacts the effectiveness of model training, it is typically necessary to clean and annotate manually. This paper proposes a deep learning-based data augmentation algorithm combined with generative adversarial network. By leveraging a small set of clean data, this approach generates additional samples to augment the dataset, thereby achieving the objective of enhancing data diversity.