In the field of data mining, conventional algorithms aren’t very suitable for data stream analysis mainly because these algorithms can not adapt to the dynamic environment of data stream mining process, and mining model and mining results can not meet the users’ actual application. To this problem, this paper presents a density granularity grid clustering algorithm to effectively accomplish the analysis task of data stream. The algorithm breaks the shackles of the traditional clustering algorithms, divides the entire mining process into off-line and on-line, and finally realizes the data stream clustering.