| The popularity of computers and mobile electronic devices,the rapid development of computing hardware and software,and the increasing participation of people in the Internet,making more and more resources available on the Internet,and how the users find the required information quickly and accurately in these complicated information is a difficult problem to be solved.The first solution to the problem of “information overload” is the search engine that can filter information.However,with the increase in the users' demand for intellectualization and personalization,the original search engine cannot provide such services,and thus a more people-oriented personalized recommendation service was born.The core of this service is personalized recommendation technology.Through the analysis of the behavior and attribute characteristics of users,and information features,the system recommends for users what they really are interested in.From the end of the 20 th century to the beginning of the 21 st century,the rise of e-commerce made the application of personalized recommendation technology more widespread.At the same time,other areas of the Internet have also introduced this technology to increase their user stickiness and gain more traffic.Through personalized recommendation services,Internet users can find useful information in a shorter time and get unexpected surprises.Similarly,the information provider can also increase the exposure of their own resource and their competitiveness,improve the conversion rate of user clicks and user loyalty.Recommendation algorithm is the key technology in personalized recommendation service.Currently,the most widely used and successful algorithm is the collaborative filtering personalized recommendation.The analysis object of this algorithm is the historical behavior of the user.However,the limited user behavior is very sparse compared to the current huge number of users and information in the Internet.This leads to the deviation of the algorithm when measuring the user's interest preference,that is,the problem of data sparsity in the recommendation algorithm.The research in this paper is based on sparse data in the recommendation system.Through in-depth study of traditional collaborative filtering algorithms,this paper proposes a personalized recommendation algorithm based on sparse data coarse graining.The main research work of this paper is as follows:(1)In order to solve the problems studied in this paper,we have an in-depth understanding of the existing recommendation algorithms and methods for solving data sparsity and focus on the most widely used collaborative filtering algorithms.At the same time,in order to solve the problem of data sparsity,a large number of data coarsening processing techniques and data reduction techniques are studied to provide theoretical support for the study of this paper.(2)The user belongings index is proposed instead of a single behavior to characterize the user's preference,which is the key to this algorithm.Since the traditional collaborative filtering measures the similarity of their interests based on the common behavior of the users,when the user behavior is sparse,may lead to inaccurate problems of the similarity calculation.On the basis of a large amount of literature reading and sorting,this paper uses the coarse grain processing technology to divide the recommended items into clusters,then calculates the user's belonging to coarse-grained clusters,and replaces the micro user's individual behavior with this macro index.The sparse scoring matrix is transformed into a dense belongings matrix,which solves the problem of data sparseness,reduces the computation of user similarity calculation.(3)Personalize recommendations for users after coarse-grained processing of sparse data.Experiments were conducted on multiple real data sets and compared with the traditional collaborative filtering to verify the pros and cons of the proposed algorithm and analyze the experimental process and results. |
PDF Full Text Request