摘要
针对协同过滤推荐算法的数据稀疏性和可扩展性问题,提出一种基于Logistic函数和用户聚类的协同过滤算法.计算用户对服务关键词的偏好度,构建用户-关键词偏好向量,并基于此向量对用户进行聚类;采用Logistic函数计算用户对服务的兴趣度,并根据兴趣度相似性在目标用户所在类内寻找其最近邻居;通过最近邻居预测用户对服务的兴趣度,将兴趣度较高的服务推荐给用户.基于真实数据集的实验证明,与传统协同过滤算法相比,本文算法能取得更高的准确率,且聚类后算法运行时间显著减少,有效地提高了推荐的实时性.
A collaborative filtering algorithm based on Logistic function and user clustering was proposed in view of the data sparsity and scalability issues of collaborative filtering.At first,a user's preference for service keywords was computed,and a user-keyword preference vector was constructed,based on which users were clustered.Then,a user's interest in service was computed by using a Logistic function.According to the similarity between users' interest,the nearest neighbors were found in the cluster,which included the target user.At last,the user's interest in a service was predicted through the neighbors' interests in the service,and services with high interest prediction were recommended to the user.The experimental results based on real data set show that this algorithm can achieve higher accuracy than traditional collaborative filtering algorithms,and the running time of clustering algorithm is significantly reduced,which effectively improves the real-time performance of recommendation.
出处
《浙江大学学报（工学版）》
EI
CAS
CSCD
北大核心
2017年第6期1252-1258,共7页
Journal of Zhejiang University：Engineering Science
基金
国家自然科学基金资助项目(61572186
61572187)
南京大学计算机软件新技术国家重点实验室资助项目(KFKT2015B04)
湖南省高校创新平台开放基金资助项目(15K043)
