Restaurant Recommendation Model In Real-time Control And Study On Cold Start Problem

With the development of Internet plus,a large number of offline service access the Internet,information overloaded,the recommendation system came into being.The recommendation system is used to solve the problem of information overload and help people to filter out the information of interest.However,only the use of recommendation system,without considering the user's current context,the results that does not meet the actual needs will be recommended to the user,this paper recommend suitable restaurants for users to solve this problem,not only consider the user's interest,but also consider the user's current location and the location of the real-time traffic information,making the travel time shorter.In the process of restaurant recommendation,it solves the problem of how to model the user's interest and how to model the real,time traffic information.On the other hand,taking into account the cold start users,the recommendation accuracy is low.Therefore the use of social networks in the relationship between friends to recommend can improve the accuracy of the recommendation.First,we use content-based recommendation algorithm and item based collaborative filtering recommendation algorithm to recommend Then,using the user's current location and the location of the real-time traffic information with the use of Baidu online real-time traffic service,we get real-time traffic calculation restaurant support,finally combined with the user the restaurant interest and real-time traffic support,support of the fusion restaurant is used to recommend,experiments show that the fusion algorithm of recommendation results average required travel time shorter,more traffic congestion,more time saved.In the semi experimental simulated data sets,the recommendation's precision and recall is slightly improved.The collaborative filtering algorithm's recommendation effect is poor,in the presence of cold start users under the condition.In the social network,we use random walk algorithm based on the relationship between project and friends,to calculate the user prediction of restaurant ratings;consider further,in reality in life,friends often go to a restaurant in knots,so the introduction of similar projects based on this kind of friends is necessary.Similarity of items because of friends is taken into account with Similarity of items because of all users.Experiment show that the improved algorithm has a slight increase in the accuracy of recommendation and recall.