Research On Personalized Hotel Recommendation System Based On User Behavior Characteristics

With the continuous development of social economic technology,the online tourism industry continues to grow and develop,and the online hotel booking market that has emerged has also developed rapidly.Many leading Internet companies have joined in,which has intensified industry competition.How to highlight the platform features and provide users with a satisfactory service has become a problem for many online hotel booking platforms.Having a personalized hotel recommendation system with good performance can solve this problem.Through reading and collecting documents,this paper finds that the classic recommendation algorithm is still the main force in the current research on hotel recommendation systems.At the same time,the implicit feedback mining for users is not enough,nor does it consider the actual application ability of the model.This article studies these issues.Because the number of users in online hotel booking platforms is usually greater than the number of hotels,calculating the hotel similarity can significantly reduce the amount of calculation.Therefore,this article uses an item-based collaborative filtering algorithm,and on this basis,the following research work is carried out:1.This article found that most of the research used the display feedback data to model the hotel recommendation system.This article first analyzed the historical scoring data of an online hotel platform in China and found that the existing research was widely used.The application's scoring data cannot reflect the user's personalized needs,so this article mainly mines and models implicit feedback data.2.However,implicit feedback data can not directly reflect user preferences and there are many implicit behaviors,so this article innovatively designed the implicit feedback preference scoring rules,and given it a new definition of calculating the hotel similarity formula.Through multiple experiments,it is proved that the implicit feedback preference scoring rule designed in this paper has a good recommendation effect,and finally an improved recommendation algorithm based on item collaborative filtering is realized.3.Considering that the user's basic characteristics will also affect the user's hotel reservation,and considering the limitations of a single algorithm and the requirements of the real-time calculation on the model,this article considers adding the XGBoost model.In this paper,the XGBoost model is used to re-learn user characteristics,hotel characteristics,and user behavior characteristics of the hotel,and a model of potential hotel reservation rate is obtained by training.This article innovatively designed a cascading model,that is,the hotel potential reservation rate model was used to filter the recommendation results of the improved item-based collaborative filtering algorithm,which made the model's recommendation effect more accurate.Finally this paper realized the use of the cascading model to build a personalized hotel recommendation system.