基于用户偏好和项目特征的协同过滤推荐算法

摘要

Using the analysis of project properties and user behavior provides the user with a valid solution recommended resources. It is recommended by the interest of the user and item’ s attributes. To the user,according to analysis of the user registration display attributes and the user’s behavior data (for browsing,viewing,downloading and sharing of project resources),as well as the quantization of the history of user behavior,the users could be divided into different neighbors. For the project,the clustering of project could divided into different resource types by its attributes and the evaluation of user to project. Therefore,the collaborative filtering algorithm is improved to solve the problems of single recommended results,little evaluation matrix data,low accuracy of recommendation as well as cold start for new users and new project. Recommended resources is achieved to change dynamically along with behavior and interest of users,to meet their requirements,achieving the purpose of the personalized recommendation,avoiding the waste of time for user to search resources in huge amounts of resources.%采用对项目属性和用户行为的分析，为用户提供了一个有效的推荐资源解决方案(通过用户的兴趣偏好和项目的属性进行推荐)。对于用户而言，根据对用户注册时的显示属性和用户的历史行为记录(对项目资源的浏览、观看、下载、分享等操作)的分析，以及对用户历史行为的量化将用户划分为不同的近邻；对于项目而言，对项目也进行相似的操作即通过项目本身具有的属性和用户对项目的评价来将项目聚类分成不同的资源类型。以此对协同过滤算法进行改进，来改善推荐结果单一、评分矩阵数据不多、推荐准确性不高以及对新用户和新项目存在的冷启动问题。实现推荐资源随用户行为、兴趣的改变而动态改变，以满足用户需求，达到个性推荐的目的，避免用户在海量资源中为搜索资源而浪费时间。