Research And Implementation On Recommendation Algorithm Based On Location And Trust Of Food Businesses

With the rapid development of the catering industry 020, more and more consumers choose the food businesses Online. However, the large numbers and huge types of Food businesses have caused serious information overload. As an effective way to solve this problem, personalized recommendation technology is widely used, but traditional food businesses recommendation based on collaborative filtering algorithm often neglects the context factors and social networks of trust relationships between users, which leads to the unsatisfactory recommendation results.In order to solve the above problems, this paper has made some researches and improvements of the food businesses recommendation algorithm based on the data from www.yelp.com, mainly including the following aspects:(1) Considering the localization features of user's activity and data sparsity, an algorithm based on location and singular value decomposition(SVD) is proposed. Firstly, the original score data is preprocessed by the reverse filling of SVD. Then, a candidate recommendation list is generated based on the score data and collaborative filtering. After that, the list is adjusted by recaculcating the recommender weight according to the travel penalty for the target user of each candidate in the list. Finally, the recommendation is made by selecting the top-n ones from the adjusted list. The experimental results on the public dataset show that the proposed algorithm has improved the recommendation accuracy.(2) To solve the low accuracy caused by the data sparsity, an alogorithm based on trust and simility is proposed by analyzing the friendship of the user in the social networks. Firstly, the direct trust between users with friendship is calculated by an improved method. Then, the indirect trust between users without friendship is obtained by the transitivity of trust. Finally, the prediction and recommendation are made based on the fusion of the trust and simility. The experimental results on the public dataset show that the proposed algorithm can alleviate the sparsity issue to some extent and get better recommendation accuracy.(3) Based on the previous two researches, this paper encapsulates the recommendation results generated by the food businesses recommendation algorithms and shares with the third-party applications. One of the key issues in this procedure is the establishment of the authentication and authorization mechanism, so an implementation method of authentication and authorization based on OAuth2.0 is proposed. And then, an experimental system built on the method verifies its effectiveness.This paper mainly focuses on improving the recommendation accuracy from the aspects of location and social networks of trust relationships between users based on traditional collaborative filtering algorithm. The experimental results show that the improved algorithms have better recommendation performance, and can provide users with more convenience by finding what they need more accurate from the massive food businesses information.