Abstract: The existing data mining applications do not allow users to create customized algorithm group on demand and can not support the usage of computing clusters well, an analysis of several components of Hadoop is made, especially focusing on Mahout which is a distributed data mining algorithm library. Then Mahout and Oozie are used to create data mining workflow in Hadoop. At last, a clustering workflow is defined and implemented. The experimental results show that the method is easy and effective for creating distributed data mining process.