Scholarlink

<p>Research on predicting booking cancellation behavior remains limited despite its operational importance. This study evaluates machine learning models based on 209,545 reservations (including 46,070 cancellations) from a four-star hotel chain (from 2015 to 2019). While XGBoost and Ridge Regression achieve high precision (97.65% and 95.65%), their high false negative rates and low recall limit sensitivity-critical applications. Conversely, KNN and a hybrid XGBoost-Ridge model excel in recall and minimize false negatives, ensuring robust cancellation prediction. Naive Bayes balances precision and recall effectively. The findings emphasize diverse metrics for model evaluation and advanced hybrid modeling for actionable insights in hotel revenue management.</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

Predicting hotel booking cancellations: a comprehensive machine learning approach

  • 期刊论文

  • 发表日期：

    2025年3月17日

查看全文
  • Apostolos Ampountolas

摘要

Research on predicting booking cancellation behavior remains limited despite its operational importance. This study evaluates machine learning models based on 209,545 reservations (including 46,070 cancellations) from a four-star hotel chain (from 2015 to 2019). While XGBoost and Ridge Regression achieve high precision (97.65% and 95.65%), their high false negative rates and low recall limit sensitivity-critical applications. Conversely, KNN and a hybrid XGBoost-Ridge model excel in recall and minimize false negatives, ensuring robust cancellation prediction. Naive Bayes balances precision and recall effectively. The findings emphasize diverse metrics for model evaluation and advanced hybrid modeling for actionable insights in hotel revenue management.