Research And Application Of Personalized Travel Recommendation Methods

Nowadays,tourism has become one of the ways for the public to relax themselves,and the rise of the Internet has provided convenience for users to obtain information on tourist attractions.However,when users choose a tourist destination,they are often faced with a large amount of scenic spot information,which makes it difficult for them to find a scenic spot that suits them and it also takes time and effort.With the widespread use of various smart mobile terminals,a large amount of user comment data was found on attractions travel websites,which can provide a wealth of available information for the construction of scenic spot recommendation.For the existing recommendation methods,although the location,category,and popularity of attractions are taken into account,the impact of users' rating habits and the popularity and viewing effect of tourist attractions which may change over time are ignored,so it is not suitable to accurately describe the different characteristics of user preferences and scenic spots.In addition,due to the sparseness of user check-in data,accurate recommendation results cannot be obtained only through the user's personal history check-in data.For this reason,inspired by the aspect-based recommendation method,various information of scenic spots and users are integrated into the recommendation model in this paper,and two improved methods are proposed.First of all,in view of the problem that the user's personal rating habits and the viewing effect of the attraction that may change over time are ignored in the existing attraction recommendation algorithms,the popularity of the attraction,the geographic location,the category of the attraction,the flow of people and the user rating are regarded as multiple aspects that affect the process of users' decision-making,and these aspects are divided into dynamic aspects that affected by time and static aspects that not affected by time.Then the impact of these aspects on the final score of attractions are respectively describe,and an attraction recommendation algorithm called TRCTE(Tourist Recommendation Considering Time Evolution)is proposed.In this method a matrix decomposition recommendation model is used to accurately describe the user's predicted scores for attractions in different periods based on user sign-in data at different times,and the accuracy of recommendation results and user experience of tourist attractions are both improved.Next,in view of the sparseness of the user's check-in data,the user's social information in the network platform is incorporated into the original recommendation model and an algorithm that integrates trusted users called TRCTU(Tourist Recommendation Considering Trusted Users)is proposed in this paper.This recommendation method not only considers the user's personal preferences,but also comprehensively considers the preferences of the user's friends.In addition,according to the user's preference for scenic spots may be related to the category of scenic spots,and users who have experienced scenic spots in the same category may have similar preferences,the category tags of scenic spots are also considered in this method when calculating the degree of trust between users,the traditional trust calculation method is improved so that the accuracy of recommendation result is improved.Finally,the dataset from Yelp website is used to verify our method.The experimental results show that the recommendation accuracy is improved and the data sparsity and cold start problems are alleviated by using the method proposed in this paper.