Research On User Portraits Of University Library Based On Reader Behavior Analysis And Multi-view Clustering Algorithm

With the development of Internet technology and the arrival of the era of big data,the convenience of accessing network resources is constantly improving.University libraries are no longer the only way for college readers to obtain literature resources.At the same time,the needs of college readers are becoming changeable.So the resources and services provided by the university library will be difficult to meet the changing needs of the readers,which causes the demand for the university library and the frequency of visiting the library is decreasing.In order to better improve the quality of service,university libraries have invested a lot of resources and have launched personalized services for college readers,but these personalized services are often based on the personal experience of librarians or the data analysis result of a certain behavior of the readers,it is concluded that the reader's changing needs cannot be captured in a timely and systematic manner with little effect.In order to assist in the university library to understand readers' needs in a more comprehensive and timely manner,this paper combines data mining technology and user portrait technology to analyze and mine readers' behavior data and user portraits.The results of the portrait provide a scientific basis for the university library to achieve accurate resource recommendation and personalized service decision-making.The main work and results of this paper are as follows:Firstly,build the reader behavior database.According to the readers behavior data accumulated in each business system and automatic system of university library,a unified database is established,and the data of each resource database is collected and stored in a unified format in the database after cleaning by ETL and other data cleaning tools.Secondly,a multi-view clustering algorithm is proposed.According to the reader's behavior data,a multi-dimensional and multi-view reader feature system is proposed.According to the constructed reader feature system,the reader can be divided into reader groups of a certain dimension or multiple dimensions,so that readers of different dimensions or multiple dimensions can be clustered,making the clustering results are more targeted.Aiming at the influence of classical k-means algorithm,which is easy to fall into local optimization and to be influenced by the attribute parameters in multi-view clustering.Good robustness,global optimization,to meet the needs of university libraries user portrait.Thirdly,a user portrait technology route based on multi-perspective clustering is proposed.The technology route use ETL technology to extract,clean and load reader behavior data from various business systems.And,according to the multi-dimensional and multi-view reader feature system proposed in this paper,the binary k-means algorithm based on Markov distance is used to cluster the reader groups after one or more dimensions are combined to obtain the clustering features of the readers,so as to obtain the user portrait.Fourthly,a university library user portrait system is designed and implemented.Based on the user portrait technology proposed in this paper,a user portrait system is implemented,through which readers can view personal information and user portrait.Readers can also view the books,services and friends recommended according to the group clustering results,so as to provide help for university libraries to achieve accurate recommendation and services.It greatly increases readers' interest and stickiness to the library.The research results of this paper have been applied in a number of universities in china,which provides a more accurate and systematic understanding of users' needs for university libraries.