基于混合特征选择算法的民航可疑订单特征提取  

Suspicious order feature selection in civil aviation based on hybrid feature selection algorithm

作　　者：丁建立[1,2] 付丽洋 曹卫东[1,2] 王家亮[1] DING Jian-li;FU Li-yang;CAO Wei-dong;WANG Jia-liang(College of Computer Science and Technology,Civil Aviation University of China,Tianjin 300300,China;Tianjin Key Lab for Advanced Signal Processing,Civil Aviation University of China,Tianjin 300300,China)

机构地区：[1]中国民航大学计算机科学与技术学院,天津300300 [2]中国民航大学天津市智能信号与图像处理重点实验室,天津300300

出　　处：《计算机工程与设计》2018年第9期2887-2893,共7页Computer Engineering and Design

基　　金：民航局创新重大专项基金项目(MHRD20150107);中国民航大学天津市智能信号与图像处理重点实验室开放基金项目(2015ASP02);中国民航大学中央高校基金项目(3122016A001;3122015C020)

摘　　要：为快速识别民航旅客订票系统中的可疑订单并及时清理占座,填补航空公司的收益漏洞,提出一种基于FilterWrapper的可疑订单特征选择算法。通过对特征的信息增益排序快速找到最优候选特征子集,利用序列前向浮动搜索算法(sequential forward floating search,SFFS)约简并提取影响可疑订单的相关维度。采用C4.5决策树算法分类建模,实验验证了可疑订单特征选择算法具有较低的计算复杂度并达到了较高的可疑订单识别率,为构建可疑订单识别模型提供了思路。To quickly identify the suspicious orders in the civil aviation booking business and timely clean up the seats to fill the revenue leakage,a suspicious order feature selection algorithm based on Filter-Wrapper method was proposed.The optimal candidate feature subset was quickly found by ranking the features using information gain.The sequential forward floating search algorithm(SFFS)was used to reduce the feature dimension of the optimal candidate feature subset and the relevant features of the suspicious order were obtained.Experiments were validated by C4.5 decision tree algorithm,and the results show that the proposed algorithm has low computational complexity and high recognition rate of suspicious order,which provides ideas for the construction of the identification model of suspicious order.

关 键 词：民航收益漏洞 可疑订单 特征选择 信息增益 序列前向浮动搜索 决策树 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]