Scholarlink

<p>This paper presents a review of advanced automated parallel test form assembly (ATA) methods for computer-based testing (CBT) using artificial intelligence approaches. Parallel test forms ensure equivalent measurement accuracy of examinees’ scores across different sets of question items and thereby enable assessments to be administered at any time and from any location. First, this paper presents a review of ATA methods, organizing them into four categories: Mixed-Integer Programming (MIP), multi-form MIP, metaheuristics, and form-maximization. Second, this paper describes a comparison of the characteristics of these methods in terms of measurement accuracy, accessibility, computational complexity, and feasible testing frequency. Moreover, based on these characteristics, the review presents an example decision framework for selecting ATA methods according to the requirements of CBT. Finally, numerical experiments were conducted for this study to compare representative ATA methods in terms of their relative strengths and weaknesses.</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

Review of automated parallel test form assembly

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年2月17日

查看全文
  • Kazuma Fuchimoto,
  • Pokpong Songmuang

摘要

This paper presents a review of advanced automated parallel test form assembly (ATA) methods for computer-based testing (CBT) using artificial intelligence approaches. Parallel test forms ensure equivalent measurement accuracy of examinees’ scores across different sets of question items and thereby enable assessments to be administered at any time and from any location. First, this paper presents a review of ATA methods, organizing them into four categories: Mixed-Integer Programming (MIP), multi-form MIP, metaheuristics, and form-maximization. Second, this paper describes a comparison of the characteristics of these methods in terms of measurement accuracy, accessibility, computational complexity, and feasible testing frequency. Moreover, based on these characteristics, the review presents an example decision framework for selecting ATA methods according to the requirements of CBT. Finally, numerical experiments were conducted for this study to compare representative ATA methods in terms of their relative strengths and weaknesses.