中文摘要：

随着网格技术的不断发展,作为网格资源管理接口的网格门户也迅速发展起来。访问网格门户的用户数和门户管理的资源数也越来越多。为了解决网格门户系统资源管理信息规模过载、服务器大规模查询和处理资源负载较高、用户获取所需资源的满意度较低等问题,该文通过分析网格门户的主要功能和特点,在结合现有协同过滤推荐算法并改进的基础上,提出了基于协同过滤的网格门户推荐模型。推荐模型包括协同过滤交互模型、处理模型和展现模型。在设计模型的基础上提出了二次组合协同推荐算法并且进行了和现有算法的比较工作。实验表明：该文提出的推荐模型可以较好地实现网格门户的个性化协同推荐功能,并且可以保证个性化推荐的准确度和质量。

英文摘要：

Grid portal is playing an important role in grid computing area.However,with the large-scale diverse resource need to be dispatched and coordinated,the portal shows its insufficient ability to deal with this complex situation and can not bear the overload of long-time transaction querying.What is worse,the users can not get their desired or expected resources.To solve these problems,grid portal recommendation architecture is presented.It consists of collaborative filter interaction layer,action layer and user render layer.In addition,a 2-way combo collaborative filter algorithm is put forward,and then the algorithm comparison is shown.Finally the experiment results improve that this architecture can be used to obtain the expected portal recommendation function and guarantee quality of personalized recommendation.