Scholarlink

<p>Determining appropriate cut-off scores for standardized language proficiency tests is critical for ensuring fairness, reliability, and validity in high-stakes decision-making. This study proposes a data-driven framework to establish optimal cut-off scores for the TOLIMO test, a standardized assessment developed and administered by the National Organization for Educational Testing (NOET) to evaluate English proficiency. Despite its significance in academic admissions and professional certification, TOLIMO has traditionally relied on expert judgment to determine pass-fail thresholds, potentially introducing subjectivity and inconsistency. To address this limitation, this research utilizes data from 461 candidates in the 189th TOLIMO exam (2021) and applies three analytical methods—Youden Index, Hurwicz Criterion, and Multilayer Perceptron (MLP) Neural Network—to derive more objective and data-driven cut-off scores. For the MLP approach, 73 features were selected using a random forest algorithm, while multiple regression analysis identified grammar, listening, and reading comprehension as significant predictors for the Youden and Hurwicz methods. The results demonstrate high accuracy across all models, with the four-layer MLP achieving the lowest root mean squared error (RMSE = 0.029) and high classification accuracy (0.975), ultimately identifying 479 as the optimal cut-off score. These findings highlight the potential of data-driven methodologies to improve the precision and consistency of cut-off score determination, thereby contributing to more equitable and evidence-based educational policies.</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

Comparative Approach to Define Cutoff Scores Using Youden Index, Hurwicz Criterion and MLP Neural Networks: TOLIMO Case Study

  • 期刊论文

  • 发表日期：

    2025年5月2日

查看全文
  • Maryam Parsaeian,
  • Ebrahim Khodaie,
  • Balal Izanloo,
  • Keyvan Salehi,
  • Sima NaghiZadeh

摘要

Determining appropriate cut-off scores for standardized language proficiency tests is critical for ensuring fairness, reliability, and validity in high-stakes decision-making. This study proposes a data-driven framework to establish optimal cut-off scores for the TOLIMO test, a standardized assessment developed and administered by the National Organization for Educational Testing (NOET) to evaluate English proficiency. Despite its significance in academic admissions and professional certification, TOLIMO has traditionally relied on expert judgment to determine pass-fail thresholds, potentially introducing subjectivity and inconsistency. To address this limitation, this research utilizes data from 461 candidates in the 189th TOLIMO exam (2021) and applies three analytical methods—Youden Index, Hurwicz Criterion, and Multilayer Perceptron (MLP) Neural Network—to derive more objective and data-driven cut-off scores. For the MLP approach, 73 features were selected using a random forest algorithm, while multiple regression analysis identified grammar, listening, and reading comprehension as significant predictors for the Youden and Hurwicz methods. The results demonstrate high accuracy across all models, with the four-layer MLP achieving the lowest root mean squared error (RMSE = 0.029) and high classification accuracy (0.975), ultimately identifying 479 as the optimal cut-off score. These findings highlight the potential of data-driven methodologies to improve the precision and consistency of cut-off score determination, thereby contributing to more equitable and evidence-based educational policies.