Based On A Novel Neighbor Selection Method For Personalized Recommendation

With the rapid development of computer and network technology, the explosive growth of information brings the information to people. At the same time, it makes people confuse about how to select the information they want. Facing the information overloading problem, recommendation system has received extensive concern. Firstly,Recommendation system analyzes the users' information, including the users' shopping habits, the items records which they have browsed and purchased,and then model with these information. Finally, forecast and recommend items that users may be interested in, and help the user to filter the majority of information which they do not concerned, and show users the valuable and interesting items. In the era of Big Data, the recommendation system is becoming more and more popular in recent years, especially among academic researchers and commercial economy researchers.Collaborative filtering recommendation system is one of the most successful methods in recommendation system. Collaborative filtering technology analysis the user's information and user's preferences to predict and recommend the favorite items for users. Collaborative filtering recommendation system includes content-based filtering, model-based collaborative filtering and hybrid collaborative filtering. Collaborative filtering recommendation system handles the contact between items and items, items and users as well as users and users, which has the same structure of network community. we can use the method which researches the community network to explore recommendation system. The article research collaborative filtering recommendation system base on items, and provide users more personalized recommendation through the combination of community-based detecting and collaborative filtering recommendation system. The main contents are as follows:(1)Introduce the recommendation system background and some kinds of recommendation system algorithm, and introduce the theory of network community detection.(2)Based on the filtering recommendation system, we proposed item-based collaborative filtering community detecting method. The contact between the items, which constitutes a structural model as well as community structure model. Analysis article structure has access to information between the articles, which will allow us to make better recommendation for users.(3) Proposed a novel neighbor selection method for personalized recommendation. In the traditional recommendation system, nearest neighbor selection method uses KNN, However, there are shortcomings and deficiencies in KNN algorithm. In the community structure, The K value of KNN algorithm is fixed, but in real network structure, K may be a changing value. Nearest neighbor collaborative filtering recommendation system makes the community structures more flexible, so that the community is more suitable for the study of the recommendation system. Therefor, it has improved the accuracy of recommendation system in some degree.