Scholarlink

<p>This study aims to predict which applicants are most likely to enrol at Higher Education Institutions (HEIs) in the Delhi-NCR region, enabling timely and data-driven admission strategies. Using 14,319 applicant records with 43 academic, economic, behavioural, and demographic attributes, the research applies advanced Machine Learning to uncover complex, non-linear patterns influencing enrolment decisions. A stacking ensemble framework combines ExtraTrees Classifier, LightGBM, Support Vector Machine, K-Nearest Neighbors, and Gaussian Naive Bayes, integrated through a Multi-Layer Perceptron as meta-classifier. Key predictors include Payment Method, Program, Domicile State, Category, Voucher Name, Discount, Origin, Publisher, Traffic Channel, and Form Completion Percentage. Model performance, evaluated via Accuracy, Precision, Recall, F1-score, Sensitivity, Receiver Operating Characteristic, and Area Under the Curve (AUC), shows a significant improvement after hyperparameter tuning. The final model achieves 85% Accuracy, 83% Sensitivity, 85% Precision, 85% F1-score, a 15% Mean Square Error, and 94% AUC, surpassing prior ensemble benchmarks. Beyond binary classification, the model generates probability scores for each applicant, ranking their likelihood of enrolment. These predictive insights empower HEIs to optimize outreach by focusing on high-probability candidates and strategically engaging potential enrollees, enhancing operational efficiency, responsiveness, and overall enrolment management through intelligent data-driven decision-making.</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

Modeling and predicting student enrolment: a meta-learning perspective

  • 期刊论文

  • 发表日期：

    2026年1月7日

查看全文
  • Simple Sharma,
  • Supriya P. Panda,
  • Seema Verma

摘要

This study aims to predict which applicants are most likely to enrol at Higher Education Institutions (HEIs) in the Delhi-NCR region, enabling timely and data-driven admission strategies. Using 14,319 applicant records with 43 academic, economic, behavioural, and demographic attributes, the research applies advanced Machine Learning to uncover complex, non-linear patterns influencing enrolment decisions. A stacking ensemble framework combines ExtraTrees Classifier, LightGBM, Support Vector Machine, K-Nearest Neighbors, and Gaussian Naive Bayes, integrated through a Multi-Layer Perceptron as meta-classifier. Key predictors include Payment Method, Program, Domicile State, Category, Voucher Name, Discount, Origin, Publisher, Traffic Channel, and Form Completion Percentage. Model performance, evaluated via Accuracy, Precision, Recall, F1-score, Sensitivity, Receiver Operating Characteristic, and Area Under the Curve (AUC), shows a significant improvement after hyperparameter tuning. The final model achieves 85% Accuracy, 83% Sensitivity, 85% Precision, 85% F1-score, a 15% Mean Square Error, and 94% AUC, surpassing prior ensemble benchmarks. Beyond binary classification, the model generates probability scores for each applicant, ranking their likelihood of enrolment. These predictive insights empower HEIs to optimize outreach by focusing on high-probability candidates and strategically engaging potential enrollees, enhancing operational efficiency, responsiveness, and overall enrolment management through intelligent data-driven decision-making.