摘要
为了实现大面积延误场景下延误航班的旅客动态成行需求量的准确预测,从旅客个体角度提出了一个两阶段预测算法框架:一是预测旅客是否会流失,二是预测会流失旅客的具体流失时间。最后通过在任意时间点上将这些旅客个体行为整合至航班的方式来实现动态成行需求量的预测。其中的流失时间预测阶段由于旅客样本流失时间分布极不均衡且实际数据具有一定模糊性等原因成为该框架下的核心难点。为了缓解由这些问题给预测效果带来的负面影响,提出了结合标记分布学习和随机森林的旅客流失时间预测模型。最后在多个真实场景数据集上进行了仿真,结果表明所提方法优于其它对比方法,实现了更加准确的旅客动态成行需求量预测。
To achieve the accurate prediction of the passenger dynamic demand of delayed flights in the large-scale delay scenario, a two-stage prediction algorithm framework from the perspective of individual passengers is proposed: firstly, predict whether the passengers will be lost;secondly, predict the specific lost time of passengers. Finally, the dynamic demand can be obtained by integrating the behavior of these individual passengers into the flight at any point in time. The lost time prediction stage is the core difficulty of the framework due to the unbalanced distribution and ambiguity of the passenger lost time. To alleviate the negative impact of these problems on the prediction effect, a prediction model of passenger lost time that combines label distribution learning and random forest is proposed. Experimental results on several real scene datasets showed that compared with multiple baseline methods, the proposed method could achieve a more accurate prediction of passenger dynamic demand.
作者
景一真
赵耀帅
傅之凤
吴格
JING Yi-zhen;ZHAO Yao-shuai;FU Zhi-feng;WU Ge(School of Computer and Information Technology,Beijing Jiaotong University,Beijing 100044,China;Beijing Key Laboratory of Traffic Data Analysis and Mining,Beijing Jiaotong University,Beijing 100044,China;Key Laboratory of Intelligent Application Technology for Civil Aviation Passenger Service,CAAC Information Network Co.,Ltd,Beijing 100105,China)
出处
《计算机仿真》
北大核心
2022年第6期26-30,46,共6页
Computer Simulation
基金
民航科技重大专项(MHRD20160109)
国家自然科学基金项目(61603028)。
