Design And Implementation Of Restaurant Recommendation System Based On User Context And Feature Information

With the rapid rise of location-based social networks and the widespread popularity of mobile intelligent terminals, users can leave their footprints and share their life experiences on relevant platforms,many of which are related to dining information. By mining the implied user preferences in these data and recommending to the user restaurants they may be interested in, opportunity is provided to resolve the contradiction between users' growing demand for dining and restaurants'weak offline promotion. To this end, this paper launched a series of research as follows:(1) Improved restaurant recommendation algorithm based on location information. First, the kernel density estimation method is introduced to establish the user's location information model. Then, a collaborative filtering recommendation model based on geographical influence and user's preference feature is proposed. Finally, a nonparametric approach is used to fuse the two models to obtain the final top-k recommendation results. Experiments on Foursquare, a public dataset, show that this algorithm can improve the recommendation accuracy and alleviate the data sparsity problem at a better recommendation efficiency.(2) Restaurant recommendation algorithm based on category labels and high order singular value decomposition. First, the tensor model is established by using the data of the restaurant's category labels. Then, the high order singular value decomposition technique is applied to decompose the tensor model. Finally, the user's predicted score for the unvisited restaurant is calculated. Experiments on the public dataset Yelp show that this algorithm can deal with the context information more reasonably and improve the recommendation accuracy(3) Restaurant recommendation algorithm based on friendship and nonnegative matrix decomposition. First, the nonnegative matrix factorization technique is used to mine the potential features of the user.Then, the similarity is calculated based on the friends' relationship.Finally, the predicted scores of the nonnegative matrix decomposition are modified by the restaurant of the user's friend with the similar preference.Experiments on the public dataset Yelp show that this algorithm can alleviate the data sparsity and improve the recommendation accuracy.(4) Collaborative filtering restaurant recommendation algorithm based on singular value decomposition feature matrix. First, the singular value decomposition technique is used to mine the user's latent feature.Then, the decomposed restaurant feature matrix is taken as the input of the item-based collaborative filtering algorithm. Finally, an improved similarity calculation method based on location information is used to measure the similarity between restaurants, and then the results of the predicted scores are generated. Experiments on the public dataset Yelp show that this algorithm can mine user's features better and improve the recommendation accuracy.Through the above research, this paper analyzes the existing problems in the restaurant recommendation field,and the restaurant recommendation system based on user context and feature information is designed and implemented. Experiments on public datasets show that the algorithm in this paper can improve the performance of restaurant recommendation system effectively.