Research And Improvement Of Collaborative Filtering Algorithm Based On Tourism Recommendation

With the development of domestic economy and the improvement of people's living standards,tourism as a leisure and relaxed lifestyle has become more and more popular.At the same time,the explosive growth of mobile internet has caused the problem of“information overload”.It is the aim for tourism recommendation system to help tourists get related information that users are interested within a large amount of data.In the traditional tourism recommendation system,it is mature to get recommendation results for users based on their geographical location and the popularity of tourist attractions.But the consumption level and travel time requirements of users are neglected.As a result,the recommendation results do not meet the consumption expectations of tourists or conflict with travel time,so it cannot meet the needs of personalized tourism services.Based on the above background,after a detailed study of domestic and foreign tourism recommendation system,this paper deeply studies the principle and process of PersonalRank algorithm,a collaborative filtering algorithm based on graph model.On the basis of PersonalRank collaborative filtering algorithm,a dynamic time weight function F(u_i)is added,and in this paper I proposed a TC-PersonalRank collaborative filtering optimization algorithm with dynamic time weight.According to the current time series of users,the algorithm get the time schedule of users,and in order to alleviate the computational pressure,the recommendation results are got by iterative calculation.In addition,the location-based social network(LBSN)is used to get the check-in data of users,and the user information obtained by the open platform of WeChat mini program is used to analyze to get the consumption data of users.And then it can be used to build user consumption model by clustering algorithm DBSCAN.Finally,the tourism WeChat mini program by TC-PersonalRank algorithm which is based on user consumption model and dynamic time weight.It is shows that the recommendation results are more suitable to satisfy the consumption level and habits of users,and meet the travel schedule of users after testing in this system.Then TC-PersonalRank algorithm and traditional recommendation algorithm are used to test in this system,the results of test show that it has a greater improvement in accuracy rate and surprise rate for the improved TC-PersonalRank algorithm.