Scholarlink

In order to timely and effectively predict the required amount of additional oil, taking a certain route of a certain airline as an example, an Extreme Gradient Boosting (XGboost) combination of adaptive boosting (Adaboost) based on combining feature screening is designed. First of all, use the gray association analysis method and random forest characteristic important analysis method to calculate the linear and non-linear correlation between the large amount of flight operation data and the additional fuel, and then perform a comprehensive feature screening. On the basis of preferred features, the established XGboost-Adaboost combination model is used to predict fuel demand as appropriate. The experimental results combined with the actual data show that the average relative error and equal square root errors of the predicted model are 5.87 and 0.2, respectively, and the prediction accuracy is higher than the prediction methods of several related types of additional fuel.

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

Recommendation Model of Additional Fuel Based on Feature Optimization and Machine Learning

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Rutong Gu,
  • Yefeng Qu,
  • Dongling Chen,
  • Wenqiang Huang,
  • Liming Deng

摘要

In order to timely and effectively predict the required amount of additional oil, taking a certain route of a certain airline as an example, an Extreme Gradient Boosting (XGboost) combination of adaptive boosting (Adaboost) based on combining feature screening is designed. First of all, use the gray association analysis method and random forest characteristic important analysis method to calculate the linear and non-linear correlation between the large amount of flight operation data and the additional fuel, and then perform a comprehensive feature screening. On the basis of preferred features, the established XGboost-Adaboost combination model is used to predict fuel demand as appropriate. The experimental results combined with the actual data show that the average relative error and equal square root errors of the predicted model are 5.87 and 0.2, respectively, and the prediction accuracy is higher than the prediction methods of several related types of additional fuel.