中文摘要：

针对传统协同过滤推荐算法评分矩阵稀疏和推荐精度不高的问题,提出一种改进的协同过滤推荐算法。通过用户属性偏好和项目流行度计算用户对项目的偏好度,结合用户平均评分对评分矩阵中未评分项目进行填充。考虑到用户兴趣随时间的变化,将基于时间的兴趣度权重函数和偏好度引入到项目相似度计算和推荐过程中,确定项目最近邻集合,从而实现最优推荐。实验结果表明,与传统协同过滤推荐算法相比,该算法较准确地反映了用户的兴趣变化趋势,并且在有效解决评分矩阵稀疏问题的同时提高了推荐准确率。

英文摘要：

For the traditional collaborative filtering recommendation algorithm meets with low recommendation accuracy and sparse matrix problem.An improved collaborative filtering recommendation algorithm is proposed.It integrates user preference on item attribute and item popularity to compute the user preference degree on item,and fills the unrated-items through the sum of user preference value and average score of users.It takes user interests change over time into account,then the time function as a weight factor in similarity calculation and recommendation process is used to find nearest neighbor set,and it achieves the optimal recommendation.Experimental result shows that this algorithm accurately reflects the change trends of user interest,and not only alleviates the sparsity matrix problem effectively,but also improves recommendation accuracy.