A Collaborative Filtering Recommendation Algorithm Based On Users Interest Drift And Interest Propagation

With the development and popularization of the Internet,the amount of information is in a way of explosive growth.Vast amounts of information not only enrich people's lives,but also becomes a challenge for the information search.Personalized recommendation system came into being.Among them,The collaborative filtering is the most classic personalized recommendation technology,this method default that user's interest is stable,that means interest does not change.However,in real life,affected by various factors outside of themselves,the user's interest is unstable,there is interest in drift.In addition,the user interaction process will affect each other's interest preferences,that interest will be spread among users.In this paper,according to the traditional collaborative filtering algorithm based on in-depth study,analyzed the shortcomings and deficiencies from previous research results,drift from interest and interest spread two aspects,mainly to do the following work:Firstly,aiming at the problem of user interest drift we proposing a drift detection algorithm based on different project types.According to project their attributes similarity calculation construction,the algorithm project clusters and based on user history access project belongs project clusters in proportion to the size to identify users of the single interest and noise.At the same time according to different project the algorithm access time dispersion degree to identify multi user interest and interest drift users.In the end,users receive the latest and the most recent interest.After this set of data pretreatment,avoiding interference with the noise of the system user data also reduces the dimension of data to a certain extent and the amount of calculation.Then,aimed at the questions between user's interest transmissions we present A fuzzy Cmeans clustering algorithm based on affine propagation.This algorithm firstly uses the message passing thought of the affinity propagation(AP)clustering and transforms the problem of interest propagation into attraction degree and the membership degree.Fuzzy CMeans(FCM)clustering is a commonly used method of data processing in collaborative filtering recommendation,the initial center of its choice directly affects the quality of clustering results.The cluster centers obtained by AP as the initial clustering center of FCM clustering,so that takes into account the interest of propagation,but also improve the quality and stability of the clustering.Finally,combining the two algorithms,we present a collaborative filtering recommendation algorithm based on Interest drift and interest propagation.In this paper,comparing different types of projects based on access time decentralized interest drift detection algorithm and the affinity propagation of fuzzy c-means clustering algorithm and the user interest drift and interest dissemination of collaborative filtering recommendation algorithm and traditional collaboration filter recommendation algorithm by experiments.Consequently,the fusion algorithm has higher accuracy and can effectively improve the quality of the system.