In　order　to　get　valuable　information,　mining　frequent　itemsets　from　multidimensional　data　stream　is　needed.　Through　introduction　of　the　concept　of　multidimensional　item　and　multidimensional　itemsets,　the　multidimensional　data　stream　is　expressed.　A　compact,　compressed　data　structure　MaxFP-Tree　is　designed　to　maintain　multidimensional　sets.　Based　on　MaxFP-Tree,　an　incremental　update　algorithm　to　mine　maximal　frequent　multidimensional　itemsets　is　given.　Experiment　results　show　that　the　model　and　the　algorithm　of　mining　multidimensional　data　streams　are　efficient.