基于多项logit模型的在线机票代理商选择行为  

Online air-ticket agent choice based on a multinomial logit model

作　　者：杨晓芳[1,2] 王喆[1] 姜海[1] 

机构地区：[1]清华大学工业工程系,北京100084 [2]内蒙古大学交通学院,呼和浩特010070

出　　处：《清华大学学报（自然科学版）》2017年第4期437-442,共6页Journal of Tsinghua University(Science and Technology)

基　　金：国家自然科学基金资助项目(71622006)

摘　　要：用户在线机票代理商选择行为的分析对旅游搜索网站机票搜索产品系统的改进非常重要。该文基于某旅游搜索网站机票搜索和预订用户的在线数据,对该网站的用户在线代理商选择行为进行分析。运用了多项logit模型理论及方法,将有关代理商的机票报价、评价、增值服务、页面位置以及市场份额的变量引入模型,并建立了可以解释该选择行为的对数价格多项logit模型。结果显示:用户的确最关注的是代理商的机票报价;搭售保险和评价人数过少都不利于用户对代理商的选择;增值服务的提供和较高的评分会增加用户对代理商的兴趣;用户更倾向于选择展示页面中靠前位置的代理商,其次是靠后的代理商;市场份额大的代理商对用户的选择行为会产生显著影响。预测检验表明该文建立的模型比历史数据预测方法有更高的预测精度。The online air ticket agent can significantly improve an air-ticket search system for online travel search engines. The paper analyzes the air-ticket agent behavior based on user air-ticket searches and booking data from an online travel search engine. A multinomial logit model is used to analyze the effects of the agents＇ quote, grades, value-added services, positions on the page and market shares to build a log-price multinomial logit model for user online air-ticket agent choices. The results show that the price quotes are the principal user concern and that including insurance and fewer raters lead to fear selections of the agent which value-added services and higher grades attract user attention. Users are more inclined to choose the agents showed on the front of the page and then the agents listed at the bottom. Agents with large market shares significantly influence the choice. Tests show that the predictions of this model are better than those using historical data.

关 键 词：多项logit模型 在线机票代理商 选择行为 

分 类 号：C93[经济管理—管理学]