Research And Implement Of Group And Event Recommendation System In Australian Activity Platform

With the rapid development of mobile Internet,more people access the Internet by their smartphone.For the weakness of location-aware,traditional recommender system cannot be applied to mobile recommendation effectively.By introducing the location-based similarity and using it with the traditional similarity,this thesis alleviates the shortcomings of the traditional recommendation calculation.Based on this,with the requirement of Australian Event Platform,this thesis propose a location-based event recommendation algorithm and a location and label based group recommendation algorithm.Finally,a prototype system using the location-based event recommendation algorithm and the location and label based group recommendation algorithm is implement.The main work of the thesis is as follows(1)Studying the related technology of the recommendation system,and the application scenarios and the advantages and disadvantages of the different recommended algorithms are analyzed.Through the analysis of two public data sets of the location-based social networks: the analysis of the Meetup data set and the Foursquare data set,it is found that the distance between the user and the activity presents the Power-Law distribution with the number of user participation activities in the distance range.Based on this,a method of user similarity calculation based on user location is proposed,and a method of calculating the similarity of position information is proposed.Combining with the relevant requirements of the recommended system in the Australian event platform,a kind of location-based event recommendation algorithm and a recommendation algorithm based on location and label are implemented respectively.Then,the two algorithms are tested and verified by the open data set.(2)To solve the problem that the location-based similarity algorithm has a lot of distance calculation,which leads to the calculation of the similarity computation too much time,this thesis transforms the arc length solving process into a rectangle diagonal solution and substitutes A very time-consuming trigonometric function in a formula.so that the calculation of similarity in the process of distance calculation of the time-consuming significantly reduced.Finally,compared with two algorithms,the result shows that the error is acceptable.At the end of this thesis,based on the proposed location-based event recommendation algorithm and location and label-based group recommendation algorithm,a prototype of the recommendation system is implemented under the overall demand of the Australian event platform.