基于旅客行为的航空旅客细分模型研究及其实现

发布时间：2018-02-13 18:32

本文关键词： 客户行为客户细分聚类 DBSCAN 核方法并行聚类　出处：《南京航空航天大学》2012年硕士论文　论文类型：学位论文

【摘要】：近年来，市场经济不稳定，民航业遭受了重大的打击。随着人们对交通需求的增大，各个交通行业都加大力度争夺市场，这无形中加重了民航业的压力。各大航空公司开始从客户的增量竞争转向客户的存量竞争，，为了把握优质客户资源，越来越多的航空公司施行了客户关系管理（CRM）系统，从而进行有效的客户细分。目前，民航领域主流旅客细分方法一般基于旅客价值，即根据旅客消费金额来实现细分。对于民航业这一特殊的领域，该方法有失偏颇，不能考虑旅客其他行为特征。因此，本文提出基于旅客行为的航空旅客细分模型，结合聚类技术进行细分。首先针对民航旅客行为属性特征，结合DBSCAN算法和核方法，消除民航旅客行为属性分布散乱、数据差异不明显等缺点。该方法是以牺牲时间效率为前提的，在信息爆炸的时代，民航客户系统中每天都会产生成千上万条信息数据，时间效率成为了必要的考虑因素。因此，本文接着利用当前较为流行的分布式并行计算技术，将该串行聚类算法并行化，计算过程中，针对民航客户“两端少，中间多”的特征，引入密度因子的概念，对合并方法进行了适当的改进。实验证明，本文的聚类方法适用于基于旅客行为的航空旅客细分模型，不仅提高了聚类结果的准确率，而且大大提升了聚类的时间效率。本文最后给出了基于旅客行为细分模型的实现案例，对A航空公司客户进行细分，并对结果进行了分析，指出相应的营销策略，为A航空公司乃至整个民航领域提出了很好的市场战略指导。
[Abstract]:In recent years, the market economy has been unstable, and the civil aviation industry has suffered a major blow. With the increasing demand for transportation, various transport industries have increased their efforts to compete for the market. This virtually aggravated the pressure on the civil aviation industry. Major airlines began to shift from incremental customer competition to customer stock competition. In order to take advantage of high-quality customer resources, more and more airlines have implemented customer relationship management (CRM) systems. In order to carry out effective customer segmentation. At present, the mainstream passenger segmentation method in civil aviation field is generally based on passenger value, that is, according to the amount of passenger consumption. For this particular area of the civil aviation industry, the method is biased and does not take into account other characteristics of passenger behavior. In this paper, an airline passenger subdivision model based on passenger behavior is proposed, which is based on clustering technology. Firstly, aiming at the characteristics of civil aviation passenger behavior attributes, combined with DBSCAN algorithm and kernel method, it eliminates the scattered distribution of civil aviation passenger behavior attributes. The method is based on sacrificing time efficiency. In the era of information explosion, thousands of information data are produced every day in civil aviation customer system, and time efficiency becomes a necessary consideration. Then, using the popular distributed parallel computing technology, this paper parallelizes the serial clustering algorithm. In the process of computing, the concept of density factor is introduced in view of the characteristics of civil aviation customers with "few ends and more intermediate". The experimental results show that the proposed clustering method is suitable for the passenger behavior based airline passenger subdivision model, which not only improves the accuracy of the clustering results, but also greatly improves the time efficiency of the clustering. At the end of this paper, an implementation case based on passenger behavior subdivision model is given, and the customers of airline A are subdivided, and the results are analyzed, and the corresponding marketing strategies are pointed out. For A airline and even the entire field of civil aviation put forward a good market strategy guidance.
【学位授予单位】：南京航空航天大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP311.13;F562;F274

【参考文献】