基于GMM的单航空公司安检区旅客到达模型  

Single airline passenger arrival model in security area based on GMM

作　　者：于之靖[1] 包邻淋 罗谦 

机构地区：[1]中国民航大学电子信息与自动化学院,天津300300 [2]中国民航局第二研究所,成都610041

出　　处：《中国民航大学学报》2017年第2期46-50,共5页Journal of Civil Aviation University of China

基　　金：国家自然科学基金项目(U1333122;U1533203);国家科技支撑计划(2012BAG04B02);中央高校基本科研业务费专项(3122014P003);四川省科技支撑计划项目(2016GZ0068);成都市战略性新兴产品研发补贴项目(2015-CP01-00158-GX)

摘　　要：运用高斯混合分布模型对某航空公司安检区旅客到达分布进行拟合分析。采用极值聚类的方法获取EM求解算法的初验分布,经过多次迭代后求得参数解。在实验环节采用某航空公司真实值机输出数据进行实验验证,结果表明GMM对单航班安检区旅客到达分布的拟合精度达到90%以上,比常用拟合方法提高了15%以上。在分峰不明显的情况下,E value-GMM与K means-GMM拟合精度相比提高了5%左右。GMM（Gaussian mixture distribution model）is used for a certain airline passenger arrival distribution fitting analysis. The extremum clustering method is adopted for the premiere distribution inspection of EM algorithm, parameter solution can be got after several iterations. In experimental part, a real airline check-in output data is adopted for experimental verification. Results show that the fitting precision of Gaussian mixture distribution is 90% or more. Compared with common fitting methods, the fitting precision is improved by more than 15%. Compared with extremum clustering and commonly used K means clustering of chi-square Gaussian mixture distribution model, the extremum clustering model has higher accuracy by about 5%.

关 键 词：单航空公司 高斯混合分布 EM算法 极值聚类 

分 类 号：V351.17[航空宇航科学与技术—人机与环境工程]