Analysis And Identification Of Online Hotel Reservation Tendency

With the continuous development of China's economy and transportation,the continuous maturity of domestic and foreign tourism market,the increasing selfawareness of tourists and the continuous change of consumption concept,the requirements for travel and living hotels are increasingly diversified,and the simple,convenient and efficient online booking method is increasingly favored by consumers.In the field of hotel online booking,consumers can book hotels through the online booking platform or directly through the hotel self built platform.At present,the orders of China's online hotel reservation industry are mainly concentrated in meituan and Ctrip.In 2019,meituan's cumulative orders accounted for more than half of the total,and Ctrip ranked second,accounting for 25%.With the popularity of Internet and smart phones,the scale of Internet users in China has become saturated.The scale of mobile Internet users in China has increased from 1.088 billion in January 2018 to 1.156 billion in December 2020.When the overall scale of Internet users reaches the upper limit,how to better explore the value of existing users,how to explore new growth points,and how to keep rising operating costs At the same time,to enhance the profits of enterprises is a topic that all companies must face in the future.In addition,the impact on COVID-19 is also the biggest impact on the hotel industry in 2020.The epidemic situation in foreign countries is still grim.The domestic customers of outbound travel can only be forced to choose to spend in the domestic market.This is also an opportunity for all online hotel platforms,and how to better after the end of the epidemic.How to let users form the habit of using and improve the number and activity of users during this period is also an important development direction.In this paper,by using the data of an online hotel booking platform in China,machine learning modeling is carried out.Compared with single model and ensemble learning,the idea of ensemble is based on random forest model.After optimizing the random forest parameters,the divided data set is modeled.The trained model is used to judge the future user decisions,and according to the random forest variables,the model is reconstructed To get the importance of the factors that affect users' decision-making,it can be used as a reference for enterprises to maximize profits from users.At the same time,it also adds logistic regression and KNN(nearest neighbor)algorithm to compare,compares the model performance through the confusion matrix and ROC curve of evaluation model standard,and finally comes to the conclusion: from the algorithm point of view,the ROC of random forest model is more accurate than logistic regression and KNN and tree model.Finally,the empirical results are summarized and analyzed.In order to improve the consumer propensity of users,we can take the strategy of logging in and getting points,accurately push users,improve the governance of the platform to improve the consumer experience of users,and give full discount coupons to price sensitive users to stimulate consumption.