Scholarlink

<p>Maintaining academic integrity during online examinations remains a critical challenge for educational institutions, despite substantial investments in research and technological advancements over the years. While progress has been made, most existing proctoring systems fall short due to a one-size-fits-all approach that fails to adapt to the different risk profiles of students during examinations. This uniform application of proctoring methods results in unnecessarily intensive monitoring for most students, leading to excessive costs for academic institutions. This research proposes a tiered proctoring framework that leverages Gaussian Mixture Models (GMMS) and changepoint detection, combined with engineered features, to stratify students into high, medium, or low-risk categories. Each student is subjected to tailored levels of proctoring based on their risk group. Validation results of the model with expert-labelled data demonstrate an impressive 94% accuracy in identifying cheating risks. Additionally, the framework is expected to provide significant cost savings by reducing the expenses of academic institutions by up to $6,250 per 1,000 low-risk students, while ensuring fairness and operational efficiency. This adaptive approach addresses financial and logistical challenges by providing scalable and sustainable proctoring solutions.</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

A Cost-Effective Tiered Proctoring Framework for Machine Learning-Based Risk Assessment of Cheating in Examination Monitoring

  • 期刊论文

  • 发表日期：

    2026年1月28日

查看全文
  • Manit Malhotra,
  • Indu Chhabra

摘要

Maintaining academic integrity during online examinations remains a critical challenge for educational institutions, despite substantial investments in research and technological advancements over the years. While progress has been made, most existing proctoring systems fall short due to a one-size-fits-all approach that fails to adapt to the different risk profiles of students during examinations. This uniform application of proctoring methods results in unnecessarily intensive monitoring for most students, leading to excessive costs for academic institutions. This research proposes a tiered proctoring framework that leverages Gaussian Mixture Models (GMMS) and changepoint detection, combined with engineered features, to stratify students into high, medium, or low-risk categories. Each student is subjected to tailored levels of proctoring based on their risk group. Validation results of the model with expert-labelled data demonstrate an impressive 94% accuracy in identifying cheating risks. Additionally, the framework is expected to provide significant cost savings by reducing the expenses of academic institutions by up to $6,250 per 1,000 low-risk students, while ensuring fairness and operational efficiency. This adaptive approach addresses financial and logistical challenges by providing scalable and sustainable proctoring solutions.