针对协同过滤算法在大数据环境下存在的可扩展性差的问题,提出一种Hadoop环境下基于改进聚类的个性化推荐算法.在Hadoop分布式计算平台上,首先在离线状态下使用基于Canopy聚类改进的模糊C均值算法构建项目聚类模型,再根据目标项目和聚类模型间相似度建立候选项目空间,最后在候选项目空间上采用基于项目的协同推荐算法在线完成推荐.实验表明,该算法在分布式集群上具有较好的可扩展性和推荐效率,且推荐精度也有所提高.%Aiming at the problem of poor scalability of collaborative filtering algorithm in the big data environment,a new personalized recommendation algorithm is proposed based on the improved clustering in Hadoop environment.On the Hadoop distributed computing platform,firstly,fuzzy C means algorithm based on Canopy clustering is used to establish the item cluster model in the offline condition,then the candidate item space is established according to the similarity between the target item and cluster model.Finally,the item-based collaborative recommendation algorithm is used to complete the recommendation on the candidate item space online.Experimental results show that the algorithm has good scalability and recommendation efficiency in the distributed cluster,and the recommendation accuracy has also been improved.
展开▼
