Scholarlink

<p>With the rapid expansion of the aviation industry, flight delays and cancellations have significantly impacted airline operational efficiency and passenger experience. To address this challenge, this study proposes a LightGBM model enhanced by &#xa0;a&#xa0;multi-strategy&#xa0;enhanced&#xa0;Golf Optimization Algorithm (MIGOA), referred to as MIGOA-LightGBM. By incorporating key techniques such as good point set initialization, dynamic centroid reverse learning&#xa0;strategy, Levy flight strategy, and differential evolution strategy, MIGOA achieves substantial improvements in global search capability and convergence speed. Comparative experiments on CEC2017 and CEC2022 benchmarks, including Wilcoxon rank-sum tests, consistently demonstrate that MIGOA outperforms GOA and other state-of-the-art optimization algorithms comprehensively. When integrated with LightGBM, the model's testing results on the Ctrip flight delay dataset reveal significant advantages in prediction metrics such as RMSE meeting the practical demands for efficient real-time forecasting in the aviation industry. This study not only provides an innovative tool for flight delay prediction but also offers a novel research perspective and methodological reference for solving complex optimization problems.</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 flight delay prediction model featuring a multi-strategy enhanced golf optimization algorithm integrated with LightGBM

  • 期刊论文

  • 发表日期：

    2025年9月15日

查看全文
  • Muyao Liu,
  • Jinlong Wang,
  • Zhihan Lyu

摘要

With the rapid expansion of the aviation industry, flight delays and cancellations have significantly impacted airline operational efficiency and passenger experience. To address this challenge, this study proposes a LightGBM model enhanced by  a multi-strategy enhanced Golf Optimization Algorithm (MIGOA), referred to as MIGOA-LightGBM. By incorporating key techniques such as good point set initialization, dynamic centroid reverse learning strategy, Levy flight strategy, and differential evolution strategy, MIGOA achieves substantial improvements in global search capability and convergence speed. Comparative experiments on CEC2017 and CEC2022 benchmarks, including Wilcoxon rank-sum tests, consistently demonstrate that MIGOA outperforms GOA and other state-of-the-art optimization algorithms comprehensively. When integrated with LightGBM, the model's testing results on the Ctrip flight delay dataset reveal significant advantages in prediction metrics such as RMSE meeting the practical demands for efficient real-time forecasting in the aviation industry. This study not only provides an innovative tool for flight delay prediction but also offers a novel research perspective and methodological reference for solving complex optimization problems.