基于IOWA算子的航班订座组合预测模型  

Combined forecasting model of flight reservation based on IOWA operator

作　　者：樊玮[1] 尚亚博 潘海莹 吴灵珊 FAN Wei;SHANG Yabo;PAN Haiying;WU Lingshan(College of Computer Science and Technology,CAUC,Tianjin 300300,China;Digital Committee of Xiamen Airlines,Xiamen 361006,Fujian,China)

机构地区：[1]中国民航大学计算机科学与技术学院,天津300300 [2]厦门航空公司数字委员会,福建厦门361006

出　　处：《中国民航大学学报》2023年第6期44-49,共6页Journal of Civil Aviation University of China

基　　金：厦门航空科技创新项目(20200618010301)。

摘　　要：航班订座预测的经典方法包括回归模型、增量模型、指数平滑模型等,在复杂多变的非线性数据环境下,各单项模型存在预测效果不稳定及健壮性不满意的情况。为解决这一问题,本文构造了一个基于诱导有序加权平均(IOWA,induced ordered weighted averaging)算子的航班订座组合预测模型,组合模型采用各单项模型的预测精度作为诱导因子,动态计算单项预测模型组合权重及构建未来的动态预测组合。但传统的IOWA组合预测模型对随机波动较大的数据适应性较差,本文在深入分析航班旅客订座量数据规律的基础上,改进了该模型在预测期的诱导因子,取得了较好的预测结果。The conventional flight reservation forecasting methods include regression model,pick-up model,exponential smoothing model and other methods.In complex and ever-changing nonlinear data environment,the prediction effect is unstable and the robustness is unsatisfactory for each single model.For this reason,this paper constructs a combined forecasting model of flight reservation based on induced ordered weighted averaging(IOWA)ope-rator.The new combined model uses the forecasting accuracy of the sub-models as induction factor,and then dynamically calculates the weights of all sub-models and constructs future dynamic prediction combination.However,the traditional IOWA combined forecasting model has poor adaptability to data with large random fluctuations.This paper improves the induction factors in the forecasting period to obtain better results based on the in-depth analysis of flight booking regular pattern.

关 键 词：航班订座 诱导有序加权平均算子 组合预测 回归模型 增量模型 指数平滑模型 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]