Improved Collaborative Filtering Recommendation Method And Its Application In Food Field

With the development of the Internet and the rapid expansion of information,Internet brings us more and more convenience,but with some trouble in the meantime.Because of the large amount of information,the traditional method to filter information is hard to help us get the accurate and effective information that we need.The information overload problem has also become one of the urgent problems to be solved today.People desperately need a kind of method that can help us get the information we need quickly and accurately.Personalized recommender system has emerged in response to this challenge and its main purpose is to provide users with different information to fit their needs and make up the weakness of the tradition method that only provides the same result for every one.Personalized recommendation has existed over 20 years.In these years,there are lots of researchers studying the recommendation system and they provide some methods to make the system accurately and effectively.Besides,they add more and more knowledge of other subjects in it.With the constant improvement and development of the personalized recommendation technology,this technology is used in more and more fields and bring us much convenience.The work of this article is to do in-deep research and analyze several commonly used recommendation algorithms,and apply the item based collaborative filtering recommendation to the field of food recommendation.To address the issues in the algorithm,this article proposes a improved item similarity calculation method with item attributes.The dataset used in this article is provided by several restaurants in the city of Changsha,we extract the food attributes and price attribute of the dishes from the original datasets,and classify the food attributes according to the national food classification system.In this paper,the traditional similarity calculation method is modified by the similarity of attribute.In the process of calculating the similarity of attribute,we use different methods according to different attributes.Finally,this paper uses different similarity correction coefficients and different numbers of nearest neighbors in the experiments.Get the best similarity correction coefficient and the nearest neighbors' number.To compare with the traditional recommendation system,the experimental results illustrate that the accuracy and the effect of the recommendation results are improved by taking the similarity of attributes into consideration.