基于潜在类别模型的航空旅客分类-《计算机技术与发展》

文章信息/Info

Author(s):: GU Zhao-jun; WANG Wei; LI Xiao-hong; College of Computer,Civil Aviation University of China

Keywords:: choice behavior; latent class probability; conditional probability; potential cluster analysis

摘要:: 潜在类别模型是用潜在的类别变量来解释外显的类别变量之间的关联,使外显变量之间的关系通过潜在类别变量来估计,进而维持其局部独立性。为研究民航旅客的选择行为偏好、改进航空公司收益管理策略,文中引进潜在类别模型,然后从PNR数据中选择合适的观察变量,概率参数化后带入模型进行建模,利用Mplus软件对模型进行求解评价,最终得到了最合理的民航旅客分类。由于是基于订票数据的,所以与以往研究相比,此方法从根本上避免了因可靠性带来的偏差风险

Abstract:: Latent class model uses the categorical latent variables to explain the connection between categorical manifest variables,makes the relationship between categorical manifest variables estimated by the categorical latent variables,and maintains its local independence.In order to study the airline passengers＇ choice behavior and improve strategy of revenue management of the airline,the latent class model for the PNR data is used to set up a model,and the Mplus software is also used to solve and evaluate,finally the best and reasonable airline passengers＇ classification is obtained.Compared with the previous researches,this method fundamentally avoids the risk of response bias