COVID-19 新冠疫情爆发以来，民航业遭受重大冲击，航空公司客运业务亏损严重。客机腹舱运力的短缺造成了航空货运价格的高涨，价格波动频繁，也给航空货运带来了机遇。航空货运基础运价的变化趋势对航空货运的收益具有重要影响，本文旨在研究航空货运基础运价的预测模型，通过分析运价的影响因素、处理和筛选运价相关的数据特征、研究基础运价预测方法，为航空货运的运价管理和货运模式的调整决策提供科学依据。本文以A航货运国际航线基础运价为研究对象，对影响航空货运基础运价的影响因素进行分析，对运价相关的数据进行统计分析和特征选择，在单个运价预测模型的基础上，用堆叠集成算法框架对预测结果进行融合，并对模型集成的方法进行改进，得到一个性能更优的航空货运基础运价预测模型。本文研究的主要结论如下：（1）本文对航空货运基础运价的影响因素进行了相关性分析。研究发现，宏观经济指标、汇率变动、民航货运市场情况、燃油价格以及A航货运的内部运营数据等都是影响运价的重要因素。通过分析这些因素与运价之间的相关性，为后续的预测模型构建提供了理论基础。（2）本文对航空货运数据分布进行了分析，考虑到航空货运数据中存在的不均衡问题，本文采用先聚类分析后分层抽样的方法对数据进行处理，以确保模型在训练过程中能够充分考虑不同类别数据的特性，从而提高运价预测模型的准确性和稳定性。同时还对货运原始数据通过降维处理得到新的货运数据特征。（3）本文比较了堆叠集成模型与随机森林回归、自适应增强算法、梯度提升回归树、极限梯度提升法在预测航空货运基础运价方面的性能。实验结果表明，本文提出的通过增强数据特征优化堆叠集成流程后的模型在预测精度和泛化能力方面都优于单个模型，决定系数比交叉验证的堆叠集成算法提升了4%，能够更准确地分析运价的变化趋势，为航空货运的基础运价预测提供了更为可靠的方法。本文的研究结论对于航空货运的动态运价管理和收益管理具有重要的实践意义，通过本文提出的基础运价预测模型，可以更好地应对航空货运市场的变化，为货运模式调整等相关决策提供科学依据，从而增加增加航空货运的收入。
Since the outbreak of COVID-19, the civil aviation industry has suffered heavy blow, and the airline passenger transport business has suffered serious losses. The shortage of belly cabin capacity in passenger planes has led to a high rise in air cargo prices, frequent price fluctuations, and also brought opportunities to air cargo. The trend of changes in basic air freight rates has a significant impact on the revenue of air freight. This thesis aims to study the prediction model of basic air freight rates. By analyzing the influencing factors of air freight rates, processing and screening data characteristics related to air freight rates, and studying basic air freight rate prediction methods, it provides scientific basis for air freight rate management and adjustment decision-making of freight modes.This thesis takes the basic air freight rates of Air Cargo International Route A as the research object, analyzes the influencing factors of the basic air freight rates, and conducts statistical analysis and feature selection on the data related to freight rates. Based on a single freight rate prediction model, a stacked ensemble algorithm framework is used to fuse the prediction results, and the method of model integration is improved to obtain a better performance air freight basic freight rate prediction model. The main conclusions of this study are as follows:(1) This thesis conducts a correlation analysis on the influencing factors of basic air freight rates. Research has found that macroeconomic indicators, exchange rate fluctuations, the civil aviation freight market situation, fuel prices, and internal operational data of Air Cargo are all important factors affecting freight rates. By analyzing the correlation between these factors and freight rates, a theoretical basis is provided for the construction of subsequent prediction models.(2) This thesis analyzes the distribution of air cargo data. Considering the imbalance problem in air cargo data, this thesis adopts a clustering analysis followed by stratified sampling method to process the data, ensuring that the model can fully consider the characteristics of different categories of data during the training process, thereby improving the accuracy and stability of the freight prediction model. At the same time, new freight data features are obtained through dimensionality reduction processing on the original freight data.(3) This thesis compares the performance of stacked ensemble models with random forest regression, adaptive enhancement algorithms, gradient boosting regression trees, and extreme gradient boosting methods in predicting basic air freight rates. The experimental results show that the model proposed in this thesis, which optimizes the stacking integration process through enhanced data features, has better prediction accuracy and generalization ability than a single model. Compared with the stack integration algorithm of cross validation, R2 has increased by 4%, which can more accurately analyze the trend of freight rate changes, providing a more reliable method for basic freight rate prediction of air cargo.The research conclusion of this thesis has important practical significance for the dynamic freight rates management and revenue management of air cargo. Through the basic fare prediction model proposed in this thesis, it can better respond to changes in the air cargo market, provide scientific basis for decision-making such as cargo mode adjustment, and thereby increase the revenue of air cargo.
