Abstract:In order to solve data sparsity and scalability of the Collaborative Filtering (CF) recommendation algorithm when the volume of the dataset is very large. After deeply analyzing the Hadoop distributed computing platform and the characteristic of Collaborative Filtering recommendation algorithm, the paper propose a optimization scheme on Hadoop platform. The experimental results show that it can effectively improve the execution efficiency of Collaborative Filtering recommendation algorithm in large data size, when it is realized by MapReduce with Hbase database on the Hadoop platform.And then, it contribute to build one recommendation system which is low cost, high-performance and dynamic scalability.