Research Of Personalised Travel Product Recommendation Based On User Browsing Behaviour

The revolution in information and communication technology on the Internet has had a profound impact on the tourism industry.In interactive tourism,tourists are increasingly inclined to use online travel websites to help them find their destinations.Travel management systems play a fundamental role in planning the perfect tour and allow system users to access relevant tour product details.Travel recommendation systems provide visitors with recommendations based on their preferences and help them to organize their leisure and travel plans.Access to relevant and accurate information is at the heart of the tourism industry and information overload has become a common phenomenon and therefore a serious problem for those seeking appropriate information.In addition,a variety of studies have been conducted on how to access information on tourism websites more effectively.The objective of this thesis is therefore to address the information overload and cold start problem,so as to accurately personalize travel products for users in corporate travel websites that are tailored to their travel needs.Based on the theoretical foundation of information foraging,this thesis closely focuses on the research theme of "personalized tourism product recommendation based on user browsing behavior",and conducts research on online user browsing behavior data.This thesis follows the research steps of theoretical research,model construction and empirical testing.Firstly,this thesis compares the research results on collaborative recommendation based on user browsing behavior and explores whether different browsing behaviors of online travel website users play a different role in the selection of travel products.Secondly,to help travel users quickly select their own personalized travel foraging information and convert potential customers into users of the recommendation system,we explore an improved model of user clustering based on information foraging,perform dimensionality reduction of user behavioral features and visualize clustering.Finally,in order to minimize the noise level to improve recommendation accuracy,relevant machine learning classification algorithms are compared and cross-validated and the interpretability of variables is increased.From the above study,some meaningful conclusions were obtained in this thesis:(1)Different information scents such as browsing behavior of users have different effects on the choice of tourism products by users' information foraging.(2)PCA-Keans clustering algorithm,which alleviates the sparsity of high-dimensional rating matrix and improves the prediction accuracy.And by matching users' ratings of tourism products and satisfaction ratings can identify the most popular tourism products in each category,thus making more accurate personalized recommendations for new users;(3)The RF algorithm in integrated classification makes the classification accuracy optimal,reduces the user cold start problem in the recommendation system and improves the quality of personalized recommendations.