Research On Recommendation Algorithm Of Fusion Of Multi-feature Information

With the rapid development of network technology,GPS positioning technology and mobile network equipment,the amount of information on the network has exploded,making it more and more difficult for people to find the information they are interested in among massive data.The recommendation algorithm effectively solves this problem and is widely applied in various fields.In location-based social networks,users can share places or experiences they are interested in with their friends or net friends by means of check-in.When users interact with location services,a large amount of space-time information and social relationship information are generated,which lays a foundation for mining user's behavioral preferences.By analyzing user's historical data and interest characteristics,location-based recommendation technology can help user's recommend places of interest to them without explicitly expressing their needs.The main research contents of this paper are as follows:Firstly,aiming at the problem of high sparsity and cold start of traditional recommendation algorithm,this paper takes into account the users' location information,time information and social relationship,and designs the recommendation algorithm of interest point-UFTL algorithm that integrates multi-feature information.In space,the similarity of location check-in and location distance is first calculated,and then the spatial similarity is calculated by linear weighted sum.In terms of time,Logistic attenuation function is introduced to analyze the influence of time factors on the change of users' interest and calculate the time similarity.In terms of social relationship,the intimacy between users and their friends is analyzed,and the similarity between friends is calculated.Finally,the similarity of the three factors is calculated by linear weighting,and the similarity of multi-feature information is obtained.Compared with other recommendation algorithms on the Foursquare data set,the results show that the UFTL algorithm designed in this paper improves the recommendation accuracy and recall rate,and effectively alleviates the problem of data sparsity.Secondly,in order to verify the rationality and practicability of the recommendation algorithm proposed in this paper,the UFTL algorithm is actually applied,and a tourist attractions recommendation system is designed and implemented.Through the overall architecture design,function module design,database design and recommendation module design of the system,the attraction recommendation APP and background management subsystem are realized.After passing the smart mobile phone test,the attraction recommendation APP can accurately recommend attractions according to user needs,thereby verifying the feasibility and effectiveness of the UFTL algorithm.