为提高民航客运量预测精准度,本文针对近18年的时间序列民航客运量数据,构建极限梯度提升树XGBoost预测模型,进行多特征分析,处理季节、节假日等主要因素,并与SVR模型进行对比。通过对比预测曲线图,反映出SVR模型在高维空间中可以找到...为提高民航客运量预测精准度,本文针对近18年的时间序列民航客运量数据,构建极限梯度提升树XGBoost预测模型,进行多特征分析,处理季节、节假日等主要因素,并与SVR模型进行对比。通过对比预测曲线图,反映出SVR模型在高维空间中可以找到最优超平面来拟合数据,XGBoost模型适用于复杂的非线性关系建模。实验结果表明,XGBoost预测模型相比于SVR向量回归模型、线性模型与随机森林模型,其精准度更高且对影响因素敏感;XGBoost模型有更高的R2和更低的MSE,能够更有效提高民航客运量的预测精度和预测稳定性,为制定航空运输生产计划和发展航空运输业提供了重要参考。In order to improve the accuracy of civil aviation passenger traffic prediction, this paper, based on the civil aviation passenger traffic data of recent 18 years, builds the ultimate gradient lift tree XGBoost prediction model, conducts multi-feature analysis, processes major factors such as seasons and holidays, and compares it with the SVR model. By comparing the prediction curves, it shows that SVR model can find the optimal hyperplane to fit the data in the high-dimensional space, and XGBoost model is suitable for complex nonlinear relationship modeling. The experimental results show that compared with SVR vector regression model, linear model and random forest model, XGBoost prediction model is more accurate and sensitive to influencing factors. XGBoost model has higher R2 and lower MSE, which can improve the forecast accuracy and stability of civil aviation passenger volume more effectively, and provide an important reference for the development of air transport production plan and air transport industry.展开更多
文摘为提高民航客运量预测精准度,本文针对近18年的时间序列民航客运量数据,构建极限梯度提升树XGBoost预测模型,进行多特征分析,处理季节、节假日等主要因素,并与SVR模型进行对比。通过对比预测曲线图,反映出SVR模型在高维空间中可以找到最优超平面来拟合数据,XGBoost模型适用于复杂的非线性关系建模。实验结果表明,XGBoost预测模型相比于SVR向量回归模型、线性模型与随机森林模型,其精准度更高且对影响因素敏感;XGBoost模型有更高的R2和更低的MSE,能够更有效提高民航客运量的预测精度和预测稳定性,为制定航空运输生产计划和发展航空运输业提供了重要参考。In order to improve the accuracy of civil aviation passenger traffic prediction, this paper, based on the civil aviation passenger traffic data of recent 18 years, builds the ultimate gradient lift tree XGBoost prediction model, conducts multi-feature analysis, processes major factors such as seasons and holidays, and compares it with the SVR model. By comparing the prediction curves, it shows that SVR model can find the optimal hyperplane to fit the data in the high-dimensional space, and XGBoost model is suitable for complex nonlinear relationship modeling. The experimental results show that compared with SVR vector regression model, linear model and random forest model, XGBoost prediction model is more accurate and sensitive to influencing factors. XGBoost model has higher R2 and lower MSE, which can improve the forecast accuracy and stability of civil aviation passenger volume more effectively, and provide an important reference for the development of air transport production plan and air transport industry.
