基于用户在线查询大数据的民航异常需求发现

发布时间：2018-03-23 01:07

本文选题：民航需求　切入点：在线机票查询　出处：《北京交通大学》2017年硕士论文　论文类型：学位论文

【摘要】：在当今民航业内,准确把握客运需求是航空公司收益管理的核心问题。随着民航市场的持续增长,对航空公司、机场、机票代理等民航相关企业的管理和运营水平尤其是市场反应能力提出了更高的要求。这就要求这些民航企业能够及时准确地掌握民航市场需求的变化,及时采取相应的市场对策,从而提高企业的运营能力和服务质量,提高收益,改善用户出行体验。在线机票预订网站上的用户查询量变化反应了民航市场需求的变化。在本文中,通过对用户在线查询行为大数据的分析,提出了一种新颖的民航异常需求发现方法,基于不同航线的用户查询量时间序列,并利用全国航线网络,从网络整体而非单条航线的视角来检测民航异常需求。首先,针对单条航线查询量时间序列上滑动窗口的异常值计算问题,提出一种多维度异常值计算的方法。分别考虑了:该窗口与本航线历史查询量曲线同期窗口子序列之间的差异、该窗口与其他航线当前查询量曲线同期窗口子序列之间的差异以及该航线查询量时间序列曲线的自身复杂性。最后通过对这三个维度计算所得的异常值相结合,得到滑动窗口子序列的初始异常值。然后,我们构建了一个由全国拥有机场城市和航线关系组成的航线网络。基于航线网络提出了一种需求异常值网络迭代优化算法,从网络整体的角度不断地迭代调整需求异常值,对上一步的异常值进行迭代修正优化,直到最后在整个航线网络上达到一种平衡,可以更为准确的得到每条航线的异常值。我们利用GDS服务商提供的真实历史查询数据集进行了实验,并与异常传统时间序列异常发现算法进行了对比实验,结果表明本文提出的方法能更有效地从用户查询行为记录中及时发现民航异常需求。最后,我们将异常需求发现方法应用在民航需求指数系统中,并利用异常发现结果对相关研究工作进行了展望。
[Abstract]:In today's civil aviation industry, accurately grasping passenger demand is the core issue of airline revenue management. With the continuous growth of the civil aviation market, airlines, airports, The management and operation level of civil aviation related enterprises, such as air ticket agents, especially the market response ability, has put forward higher requirements. This requires these civil aviation enterprises to grasp the changes in civil aviation market demand in a timely and accurate manner. Timely adoption of corresponding market countermeasures to improve the operation capacity and service quality of the enterprise, and improve the revenue, Improve the user travel experience. The change of user query volume on the online ticket booking website reflects the change of the civil aviation market demand. In this paper, through the analysis of the user online query behavior big data, In this paper, a novel method of detecting civil aviation abnormal demand is proposed, which is based on the time series of user query of different routes, and uses the national airline network to detect the abnormal demand of civil aviation from the perspective of the network as a whole rather than a single route. In order to solve the problem of calculating the outliers of sliding window on the time series of query quantity on a single route, a method of calculating outliers in multiple dimensions is proposed. The differences between this window and the sub-sequences of the synchronization window of the historical query curve of this route are considered, respectively. The difference between this window and the current query volume curve of other routes and the complexity of the time series curve of the query volume of this route are also discussed. Finally, the outliers calculated from these three dimensions are combined. The initial outliers of the sliding window sub-sequences are obtained. Then, we construct a route network composed of cities and routes owned by the whole country. Based on the airline network, an iterative optimization algorithm for outliers network is proposed. From the point of view of the whole network, the demand outliers are adjusted iteratively, and the outliers in the previous step are modified and optimized iteratively until a balance is reached in the whole airline network. We make use of the real history query data set provided by GDS service provider to carry on the experiment, and compare it with the anomaly discovery algorithm of the traditional time series of anomaly, we can get the outlier value of each route more accurately. The results show that the method proposed in this paper can more effectively detect the abnormal requirements of civil aviation from the user's query behavior record. Finally, we apply the anomaly requirement discovery method to the civil aviation demand index system. The results of anomaly detection are used to forecast the related research work.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F562;TP311.13

【相似文献】