摘要
针对传统的协同过滤算法中存在数据稀疏性和冷启动的不足,分析目前已有的解决方案,提出基于用户兴趣传播的协同过滤算法。在改进算法中可以让用户兴趣进行直接传播,使得用户兴趣游走以及更新,计算用户-兴趣的分布矩阵从而获取用户兴趣的相似性,然后对上述过程通过算法描述其实现过程,最后对算法进行实验分析。在这个算法当中不仅解决用户的兴趣的相似性计算问题,又考虑到其他的邻居的用户对于目标用户兴趣偏好的影响,在数据稀疏性的情况下保证了算法的有效性,在一定程度上提高了性能。仿真实验表明,算法的性能具有可行性和有效性。
In view of the deficiencies of data sparsity and cold start the traditional collaborative filtering algorithm has,we analyse current existing solutions,and put forward the user interest propagation-based collaborative filtering algorithm.In improved algorithm,user interest is allowed to directly propagate,this makes the user interest wander and update.We calculate user-interest distribution matrix so as to obtain the similarity of user interest,and then describe through algorithm the implementation process of the above procedure.Finally we analyse the ex-periment of the algorithm.In this algorithm,besides solving the similarity computation problem in regard to user interest,it also considers the influence of other neighbouring users on the interest preference of target users,and ensures the effectiveness of the algorithm in sparse data situation,improves the performance to certain extent.Simulation experiment show that the performance of the proposed algorithm is feasible and effective.
出处
《计算机应用与软件》
CSCD
2015年第10期95-100,104,共7页
Computer Applications and Software
基金
广东省计算机网络重点实验室开放基金项目(CCNL200709)
关键词
数据稀疏性
用户兴趣
直接传播
兴趣游走
兴趣偏好
协同过滤
Data sparsity
User interest
Direct propagation Interest wandering
Interest preference
Collaborative filtering
