Collaborative Filtering Recommendation Based On Weighted Networks

With the rapid development of science and technology, the Internet and othernetworks around the world into an exponential growth available to us more and moreinformation resources. We often need some confidence in filtering technology, fromthe huge pool of resources, rapid access to the most relevant information. Collaborativefiltering has been successfully applied to the Internet as a more mature recommendedtechnology scene. It is mainly based on the property or interest in a similar userexperience and suggestions as to provide personalized recommendation on the basis ofThrough collaborative filtering to help collect with similar preferences or attributes ofusers, and their views to other users in the same network as a reference, and is alsoconsistent with the mental people are usually accustomed to before thedecision-making reference to the views of others.Recommended in theory and practice have been the rapid development, how totake full advantage of an effective network built by users and product information,provide users with accurate and diverse recommended results became the core ofpersonalized recommendation technology development. In this paper, we mainly studybased on the weighted network of personalized recommendation technology, weightednetwork is the recommended network of relationships established by the user set andproduct set with each other. Weighted network can be seen everywhere in practicalapplications, B2C Web site users to buy goods of scoring records, purchase time, timeto market, the level of trust between users, users to add tag information, etc. can beused as the weights of the network information to participate in recommended, so ithas a high research value.Traditional referral model to consider the original scoring, we propose toeliminate external factors based on the purchase of interval weights of the benchmarkmodel for evaluation of information. We studied the right to add the value in the taginformation, the recommended network model, and tag information, create a newweight model, changing the similarity calculation program to improve the accuracy ofthe recommended. Personalized recommendation system, user information may be leaked. We address these issues focus on collaborative filtering, privacy protectionmodel, followed by a new information protection strategy to protect users’ privacyinformation system filtering-based recommender system.