The Research Of One-way Flight Search Sort Based On Discrete Choices Model

Recently,people can finish airline search and reservations through kinds of travel Web site.After users choose depart place,destination and initial time and do the search,the result pages come out.Users could pick the travel fit for themselves to reserve.We research users'behavioral traits when they choosing their travel,using the data of one travel Web site,and we build a quantitative model which could explain users' decision.Based on the research result,we do the research of one-way travel search rank further,and then we have got the rank equation,which could be applied to the practical account.Our research could provide a reference to the improvement of travel Web site on the search result ranking.First,the article states the character,principle and development of discrete choices model.In the introduction,we emphasize the Log-it discrete choices model,on which our research about users' choices to make flight reservations is based.Meanwhile,we give a brief generalization of the development of search engine and sorting algorithm.Then we study the behavior of users in searching and booking tickets,applying the data of one tourist searching website.Based on the Log-it discrete choices model,taking four cities,Beijing,Shanghai,Guangzhou and Chengdu as the example,we put the flight attributes displayed in the searching result pages into the MNL and NL model.After judgment and appraisal,we've got the model which could explain the choices of users to flights.The analyzing result shows that the price,the departure time,plane models,flight sorts,airlines,display position and other factors could influence users' choices.And then,we've applied the model,which has been set before,to build the discrete choices model of flight search and rank optimization.Firstly,we give 11 factors which may influence users' choices.Secondly,we replace them by 9 variances after principal component analysis.Thirdly,applying the 14-day data provided by the website,we do the multiple linear regression and get the rank equation.Finally,we have done the significance tests and the result of testing shows that the fitting parameters and quantitative evaluation are stable.In the end,the article discusses the universality of our model from 3 parts:flight line division,life cycle of the rank equation,and the testing sample size.In addition,we've given strategies how we adjusts the discrete choices model in different situations,to guarantee the rank equation we've got from the model could be still suitable.