中文摘要：

针对传统协同过滤算法不能及时反应用户的兴趣变化、时效性不足而导致推荐精度不高的问题,提出一种基于用户兴趣和项目属性的协同过滤算法。在传统协同过滤基础上综合考虑评分时间、相似度以及项目属性等因素,首先在计算相似度过程中加入基于时间的用户兴趣度权重函数,然后再与项目属性相似度进行融合,最后进行项目预测与推荐。在Movielens数据集上的实验结果表明,所提出的算法与已有的经典算法相比,平均绝对误差降低了3%-6%,有效提高了推荐的准确性。

英文摘要：

Aiming at the problem that traditional collaborative filtering algorithm can , t response to user, s interestchanges timely , lack of timeliness leads to recommend accuracy is not high , a collaboratuser interest and item properties is proposed. On the basis of the traditional colof scoring time , similarity and item properties , first we add the user preference time weight function to the process of computing similarity , then merge with the similarity of tem properties , finally make the item prediction and recommendation. Compared with the newly proposed algorithm and the existing algorithm , theMovielens demonstrate that the MAE （Mean Absolute Error） reduces by 3% ？6% ,which effectively improves therecommendation accuracy.