基于网络搜索数据的区域旅游指数及其微观动态:以西安为例

The Construction of Regional Tourism Index and Its Micro-Dynamic Characteristics:A Case Study of Xi’an

地理位置与交通条件是旅游的重要先决条件,然而高峰期突然蜂拥而至的客流往往令身陷其中的游客多有抱怨和排斥,景区管理部门亦力有不逮,因此,如何及时预警、调度和配置有限的旅游资源便成为大众和旅游管理部门的热点话题.最直接的方法是客流量预测,然而目前的客流量统计主要来源于对旅行社、景区及其周边酒店等机构部门的事后统计和推断,这类数据需要层层汇总上报审批,发布相对滞后,并且常以月度或季度为频率.网络搜索数据记录了游客成行之前的需求与偏好,数据本身是前置的、实时的,而且是直接对游客行为痕迹的记录,数据质量也不依赖于被调查一方的动机和配合程度.以古丝绸之路的起点西安为例,基于用户的旅游信息搜索行为,通过设置初始关键词和拓展关键词筛选相关搜索词构造区域旅游指数,采用HEGY季节协整检验和基于X12季节调整的混合模型均发现,区域旅游指数并不具有显著的前兆效应,只是与实际旅游客流量存在同期关联,可以进行实时预测.进一步,在Prophet预测模型中引入节假日效应,显著地降低了拟合与预测误差.

Geographical location and traffic condition are important aspects for tourism, but the sudden influx of tourists often makes people feel crowded and querulous, to which scenic managements are helpless. Therefore, it is a hot topic of public and tourism management about how to timely warn, schedule and allocate limited tourism resources. The most direct way is to forecast tourist flow, but data of current tourism statistics and inferences are mainly from the travel agency, scenic area,surrounding hotels and other departments afterwards. These data need layers of summary and submission, and they are usually released by month or quarter with a time lag. Advances of information technology have given rise to a massive amount of big data generated by users. In particular, web search data records the needs and preferences of tourists before the trip, which is advanced and real-time record of tourists＇ behavior and does not depend on respondents＇ motivation and degree of coordination. This paper takes Xi＇an as an example, which is the starting point of the ancient Silk Road and popular destination for humanistic tourism. By choosing and filtering basic search queries and their related searches, this paper first constructs the regional tourism index. Then the relationship between index and tourist flow is studied. The checkout of HEGY seasonal cointegration test and the mixed model based on X12 seasonal adjustment shows that index has association with tourist flow over the same time, which can be used for nowcast. Furthermore, the paper introduces holiday effect into the Prophet prediction model, significantly reduces the error of fitting and prediction.