Scholarlink

<p>To address the bias risk in digital literature resources and the inefficiency of manual review in library automation systems, this study proposes an automated bias risk assessment system that integrates Term Frequency-Inverse Document Frequency (TF-IDF) and Support Vector Machine (SVM) models. The approach combines TF-IDF and BERT for feature extraction and vectorization, employs SVM for classification and assessment, and introduces the Whale Optimization Algorithm (WOA) to optimize SVM parameters. The experiment showed that the shortest training time of this method reached 1.7&#xa0;s, and the highest weighted F1 score reached 93.4%. In the practical application of the evaluation system, the area under the curve of the research method reached 0.863, which was the closest to 1 numerically. The performance fluctuation was the smallest at 2.2%, and the false alarm rate was the lowest at 3.7%. It demonstrates stronger classification performance and robustness, promoting the development of intelligent and personalized risk assessment systems for libraries.</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

TF-IDF joint SVM model in library automation bias risk assessment system

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年2月18日

查看全文
  • Yu Zhao,
  • Ying Che

摘要

To address the bias risk in digital literature resources and the inefficiency of manual review in library automation systems, this study proposes an automated bias risk assessment system that integrates Term Frequency-Inverse Document Frequency (TF-IDF) and Support Vector Machine (SVM) models. The approach combines TF-IDF and BERT for feature extraction and vectorization, employs SVM for classification and assessment, and introduces the Whale Optimization Algorithm (WOA) to optimize SVM parameters. The experiment showed that the shortest training time of this method reached 1.7 s, and the highest weighted F1 score reached 93.4%. In the practical application of the evaluation system, the area under the curve of the research method reached 0.863, which was the closest to 1 numerically. The performance fluctuation was the smallest at 2.2%, and the false alarm rate was the lowest at 3.7%. It demonstrates stronger classification performance and robustness, promoting the development of intelligent and personalized risk assessment systems for libraries.