摘要
航班需求预测是航空公司收益管理的关键技术。BP神经网络用大量的历史数据进行学习,能够记忆复杂的历史订座规律和销售趋势,提出了一种基于BP神经网络的航班需求预测模型。通过对历史数据进行主成分分析获得该模型,用一元回归法和相关系数法对训练质量进行评估,对模型作了置信区间分析。将该模型与增量法、回归法进行了对比,具有在线预测速度快、预测精度相对较高等优点。
Fight demand forecasting is the core technique for airline revenue management. Back Propagation Neural Network is trained by a great deal of historical data,and it can memory the complicated rules and sale trend.So a new flight demand forecasting model based on Back Propagation Neural Network is presented.The model is acquired by Primary Component Analysis of history data,and analyzed by Believed Trivial Analysis.The training result of BP is evaluated by Linearity Regression and Correlation Coefficient Methods.Compared with widely used pick-up and Regression algorithms,the model is more faster and accurate.
出处
《中国民航学院学报》
2004年第6期44-49,共6页
Journal of Civil Aviation University of China
基金
中国民航总局科技基金项目(2002-18).
关键词
航班需求预测
BP神经网络
收益管理
主成分分析
demand forecasting
back propagation neural network
revenue management
primary component analysis
