摘要
针对传统协同过滤推荐算法模型过于粗糙和推荐精度较低的问题,提出了一种新的基于用户量化属性的多维相似度的协同过滤推荐算法.该算法考虑到个别项目对相似度计算的影响,利用最大差值特性进一步描述用户评分相似度,并结合用户量化属性,构建用户兴趣偏好模型,依此阐明了新的相似度计算方法,利用该方法获取目标用户的近邻用户和预测评分,最终实施推荐.实验结果表明该算法可以有效的提高推荐质量.
Considering the fact that the traditional collaborative filtering recommendation algorithm is rough and has low recommendation accuracy, a new recommendation algorithm is proposed based on multi-similarity of user quantitative attributes. Taking the impact of individual items into account, the algorithm describes the user rating similarity by using the maximum difference feature, and combines the user quantitative attributes to build the user interest preference model, thus the new similarity calculation method is elaborated. The neighbor user of the target user and the prediction score are obtained by the method, which contributes to the effective recommendations. The results show that the proposed algorithm can effectively enhance the quality of recommendations.
出处
《江西理工大学学报》
CAS
2017年第3期86-91,共6页
Journal of Jiangxi University of Science and Technology
基金
江西省社科规划项目(13YD020)
关键词
推荐算法
协同过滤推荐
最大差值
量化属性
近邻用户
recommendation algorithms
collaborative filtering
maximum difference
quantitative attribute
neighbor user
