Abstract: Internet is a huge and widely distributed information service center, the vast amounts of data generated on the Internet arc usually geographically distributed, heterogeneous, dynamic and become more complex, it can not meet the requirements if we use the existing centralized data mining methods. To solve these problems,proposed a cloud computing-based Web data mining method, the massive data and mining tasks will be decomposed on multiple computers parallely processed. We use open platform-Hadoop to establish a parallel association rules mining algorithm based on Apriori, and it tests and veriftes the efficiency of system. This paper proposed a design thinking that "migrate the calculation to the store"，the calculation will be implemented on the local storage nodes, thus it can avoid the large amount of data transmission on the network, and will not take a lot of bandwidth.

Key words: Cloud-computing, Data mining, Map/Reduce, Association rules