NERMS中个性化资源推荐的设计与实现
摘要
随着计算机技术和网络技术的发展与成熟以及各国对教育重视程度的增强，网络教育在世界各国日益普及。获取信息的途径已由教室、实验室和图书馆，扩展到因特网所覆盖的任何场所。然而，网络只是信息传播的载体，信息资源的汲取和共享才是人们使用网络的目的所在。要发挥网络教育区别于传统教育的优势，就必须有丰富的教学信息资源支持，同时还要能对这些信息进行有效的组织和管理。
网络教育资源管理系统NERMS（Network Educational Resource Management System）是我们承担的吉林省科学技术厅的重大项目。NERMS的主要目标是对繁多的网络教育资源进行有效的组织和管理，以便于网络教育资源的高度共享和便利获取，从而加快网络教育资源的开发和促进网络教育的发展。
新一代的信息服务将是主动的个性化信息服务，如何从海量的数据和信息中高效地获取有用知识，如何从动态变化的信息中及时地获取最新信息，如何提高信息检索与推送的智能水平，以及如何满足各种用户不同的个性化需求等，都是新的信息服务系统面临的挑战性课题。
系统为每个用户提供了一个个性化网页，以满足用户个性化的服务需求。每个用户可以选择自己的个性化网页中的内容，个性化网页中的一个重要内容是系统自动地为用户推荐的资源的列表，系统为用户推荐资源的依据是各个用户的兴趣爱好及行为特征，该功能为用户在大量数据中查找感兴趣的资源提供了快捷方便的途径。
本文首先介绍了个性化推荐技术，然后把该技术应用到了NERMS中，开发出了个性化网页中的资源推荐功能，并详细介绍了具体的设计和实现过程。本文中为用户推荐资源的主要依据是用户与资源的交互记录，包括浏览资源、收藏资源及下载资源。通过对这些交互信息的分析，找到用户的兴趣爱好所在，近而找到与该用户兴趣相近的用户，以此为用户形成推荐。
本文采用了基于协作式过滤的推荐和聚类相结合的技术，按照先聚
类、再找近邻、最后形成推荐的步骤来为用户推荐资源。对于每一步操作所用的信息，也都进行了降维和稠密化处理，并选择最佳时机启动每项操作，尽量提高系统推荐的准确性和程序运行的效率。
本文采用多层体系结构，使用MVC模式，运用J2EE技术来完成系统的开发，并使用IBM公司的DB2作为后台数据库管理系统，IBM WebSphere Studio Application Developer作为开发测试环境，WebSphere Application Server作为后台应用服务器。整个开发过程思路清晰，层次分明，调试灵活。系统在测试环境下，试运行三个月，经实验证明推荐效果较好。
With the development and maturity of computer and network technology, and the growth of educational attention degree of various countries, network education is popularized day by day in countries all over the world. The way to obtain information is expanded from classroom, laboratory and library to any place that Internet covers. However, the network is only the information carrier, the drawn and sharing of information resources is the purpose of people using the network .To take advantage of network education, there must be abundant network resources and we must can organize and manage the resources effectively at the same time.
NERMS (Network Educational Resource Management System) is a great project sponsored by Science Committee of JiLin Province and assigned to Knowledge Engineering Lab of the Institute of Computer Science and Technology in JiLin University. The aim of the project is to organize and manage various kinds of educational resources effectively so that people can share and gain them efficiently and can increase the speed of developing network education .
New information service is individual and initiative .How to gain useful knowledge efficiently from lots of data and information ,how to gain the newest information immediately from rapid exploding information ,how to increase the level of retrieving and pushing the information intelligently ,and how to satisfy all kinds of individual requires of users etc ,are the challenge subject of new information service system will face with .
In this system ,we provide an individual page for every user to satisfy their individual requirements .Every user has the right of choosing the content of his individual page .An important part of individual page is the resource list which is provided automatically by the system according to everyone’s interest and action character .This function provides an immediate
and convenient approach for users to find their own fond resources in all kinds of resources.
This paper firstly describes the technique for making personal recommendations .Secondly it applies the technique to NERMS ,and develops a personalized page which can automatically provide resources that every user is fond of .At last ,we discuss the design and implementation of the function .We providing resources for users are according to the track records of users and the system ,which include the browsed resources, the downloaded resources and the collected resources .Through the analysis of the track records ,we can gain the interest of the user ,and find the neighbors for him ,so that we can recommend resources to him.
We have used CF-based recommendation and clustering technique . There are three steps to generate recommendations .We firstly classify all users with clustering arithmetic ,secondly we find neighbors for every users ,at last we generate recommendations for every user .The information of each step requires is dealt with Dimension Reduction and Densification, and we choose the best opportunity to start each operation to increase the efficiency of the system .
We have chosen the multilayer structure and used MVC model and J2EE technology in developing the system .We use DB2 of the IBM company as the background database management system ,and IBM WebSphere Studio Application Developer as the development and testing environment ,and IBM WebSphere Application Server as the background server .In the test environment ,the system has be running for three months . The process of the development is clarity and agility .It is proved that the result of recommendations is good .
网络教育资源管理系统NERMS（Network Educational Resource Management System）是我们承担的吉林省科学技术厅的重大项目。NERMS的主要目标是对繁多的网络教育资源进行有效的组织和管理，以便于网络教育资源的高度共享和便利获取，从而加快网络教育资源的开发和促进网络教育的发展。
新一代的信息服务将是主动的个性化信息服务，如何从海量的数据和信息中高效地获取有用知识，如何从动态变化的信息中及时地获取最新信息，如何提高信息检索与推送的智能水平，以及如何满足各种用户不同的个性化需求等，都是新的信息服务系统面临的挑战性课题。
系统为每个用户提供了一个个性化网页，以满足用户个性化的服务需求。每个用户可以选择自己的个性化网页中的内容，个性化网页中的一个重要内容是系统自动地为用户推荐的资源的列表，系统为用户推荐资源的依据是各个用户的兴趣爱好及行为特征，该功能为用户在大量数据中查找感兴趣的资源提供了快捷方便的途径。
本文首先介绍了个性化推荐技术，然后把该技术应用到了NERMS中，开发出了个性化网页中的资源推荐功能，并详细介绍了具体的设计和实现过程。本文中为用户推荐资源的主要依据是用户与资源的交互记录，包括浏览资源、收藏资源及下载资源。通过对这些交互信息的分析，找到用户的兴趣爱好所在，近而找到与该用户兴趣相近的用户，以此为用户形成推荐。
本文采用了基于协作式过滤的推荐和聚类相结合的技术，按照先聚
类、再找近邻、最后形成推荐的步骤来为用户推荐资源。对于每一步操作所用的信息，也都进行了降维和稠密化处理，并选择最佳时机启动每项操作，尽量提高系统推荐的准确性和程序运行的效率。
本文采用多层体系结构，使用MVC模式，运用J2EE技术来完成系统的开发，并使用IBM公司的DB2作为后台数据库管理系统，IBM WebSphere Studio Application Developer作为开发测试环境，WebSphere Application Server作为后台应用服务器。整个开发过程思路清晰，层次分明，调试灵活。系统在测试环境下，试运行三个月，经实验证明推荐效果较好。
With the development and maturity of computer and network technology, and the growth of educational attention degree of various countries, network education is popularized day by day in countries all over the world. The way to obtain information is expanded from classroom, laboratory and library to any place that Internet covers. However, the network is only the information carrier, the drawn and sharing of information resources is the purpose of people using the network .To take advantage of network education, there must be abundant network resources and we must can organize and manage the resources effectively at the same time.
NERMS (Network Educational Resource Management System) is a great project sponsored by Science Committee of JiLin Province and assigned to Knowledge Engineering Lab of the Institute of Computer Science and Technology in JiLin University. The aim of the project is to organize and manage various kinds of educational resources effectively so that people can share and gain them efficiently and can increase the speed of developing network education .
New information service is individual and initiative .How to gain useful knowledge efficiently from lots of data and information ,how to gain the newest information immediately from rapid exploding information ,how to increase the level of retrieving and pushing the information intelligently ,and how to satisfy all kinds of individual requires of users etc ,are the challenge subject of new information service system will face with .
In this system ,we provide an individual page for every user to satisfy their individual requirements .Every user has the right of choosing the content of his individual page .An important part of individual page is the resource list which is provided automatically by the system according to everyone’s interest and action character .This function provides an immediate
and convenient approach for users to find their own fond resources in all kinds of resources.
This paper firstly describes the technique for making personal recommendations .Secondly it applies the technique to NERMS ,and develops a personalized page which can automatically provide resources that every user is fond of .At last ,we discuss the design and implementation of the function .We providing resources for users are according to the track records of users and the system ,which include the browsed resources, the downloaded resources and the collected resources .Through the analysis of the track records ,we can gain the interest of the user ,and find the neighbors for him ,so that we can recommend resources to him.
We have used CF-based recommendation and clustering technique . There are three steps to generate recommendations .We firstly classify all users with clustering arithmetic ,secondly we find neighbors for every users ,at last we generate recommendations for every user .The information of each step requires is dealt with Dimension Reduction and Densification, and we choose the best opportunity to start each operation to increase the efficiency of the system .
We have chosen the multilayer structure and used MVC model and J2EE technology in developing the system .We use DB2 of the IBM company as the background database management system ,and IBM WebSphere Studio Application Developer as the development and testing environment ,and IBM WebSphere Application Server as the background server .In the test environment ,the system has be running for three months . The process of the development is clarity and agility .It is proved that the result of recommendations is good .
引文
Paul Resnick and Hal R.Varian. Recommender Systems. March 1997/Vol. 40, No. 3 COMMUNICATIONS OF THE ACM, 56-58
Akira Sato, Takahisa Ando, Hiroya Inakoshi, and Seishi Okamoto. Personalization System based on Dynamic Learning.
John S.Breese, Davia Heckerman, and Carl Kadie. Emperical Analysis Of Predictive Algorithms For Collaborative Filtering. Proceedings of the Fourteenth Conference on University in Artificial Intelligence, Madison, WI, July, 1998.
Gediminas Adomavicius and Alexander Tuzhilin. Extending Recommender Systems: A Multidimensional Approach.
Badrul W.Sarwar. Sparsity Scalability and Distribution in Recommender Systems.
Badrul Sarwar, George Karypis, Joseph Konstan, and John Riedl.Analysis of Recommendation Algorithms for E-Commerce. EC’00,October 17-20, 2000, Minneapolis, Minnesota. Copyright 2000 ACM 1-58113-272-7/00/0010
J. Ben Schafer, Joseph Konstan, and John Riedl. Recommender Systems in E-Commerce.
Vijay S.Iyengar and Tong Zhang:Empirical Study of Recommender System Using Linear Classifiers.
Weiyang Lin, Sergio A Alvarez and Carolina Ruiz. Collaborative Recommendation Via Adaptive Association Rule Mining.
S. Guha, R. Rastogi, and K. Shim, "CURE: An efficient clustering algorithm for large databases". In Proceedings of ACM SIGMOD International Conference on Management of Data, pages 73--84, New York, 1998.
T. Zhang, R. Ramakrishnan, and M. Livny. "Birch: An efficient data clustering method for very large databases". In Proceedings International SIGMOD Conference on Management of Data (SIGMOD 199
Badrul Sarwar, George Karypis, Joseph Konstan, and John Riedl .Analysis of Recommendation Algorithms for E-Commerce. EC’00, October 17-20, 2000, Minneapolis, Minnesota. Copyright 2000 ACM 1-58113-272-7/00/0010.
Tong Sun and Andr′e Trudel. An implemented e-commerce shopping system which makes personal recommendations.
Akira Sato, Takahisa Ando, Hiroya Inakoshi, and Seishi Okamoto. Personalization System based on Dynamic Learning.
John S.Breese, Davia Heckerman, and Carl Kadie. Emperical Analysis Of Predictive Algorithms For Collaborative Filtering. Proceedings of the Fourteenth Conference on University in Artificial Intelligence, Madison, WI, July, 1998.
Gediminas Adomavicius and Alexander Tuzhilin. Extending Recommender Systems: A Multidimensional Approach.
Badrul W.Sarwar. Sparsity Scalability and Distribution in Recommender Systems.
Badrul Sarwar, George Karypis, Joseph Konstan, and John Riedl.Analysis of Recommendation Algorithms for E-Commerce. EC’00,October 17-20, 2000, Minneapolis, Minnesota. Copyright 2000 ACM 1-58113-272-7/00/0010
J. Ben Schafer, Joseph Konstan, and John Riedl. Recommender Systems in E-Commerce.
Vijay S.Iyengar and Tong Zhang:Empirical Study of Recommender System Using Linear Classifiers.
Weiyang Lin, Sergio A Alvarez and Carolina Ruiz. Collaborative Recommendation Via Adaptive Association Rule Mining.
S. Guha, R. Rastogi, and K. Shim, "CURE: An efficient clustering algorithm for large databases". In Proceedings of ACM SIGMOD International Conference on Management of Data, pages 73--84, New York, 1998.
T. Zhang, R. Ramakrishnan, and M. Livny. "Birch: An efficient data clustering method for very large databases". In Proceedings International SIGMOD Conference on Management of Data (SIGMOD 199
Badrul Sarwar, George Karypis, Joseph Konstan, and John Riedl .Analysis of Recommendation Algorithms for E-Commerce. EC’00, October 17-20, 2000, Minneapolis, Minnesota. Copyright 2000 ACM 1-58113-272-7/00/0010.
Tong Sun and Andr′e Trudel. An implemented e-commerce shopping system which makes personal recommendations.