This　paper　proposes　an　algorithm　for　clustering　XML　data　stream　using　sliding　window.　It　is　a　dynamic　clustering　algorithm　based　on　XML　structure.　Firstly,　we　use　level　structure　to　represent　XML　document,　which　is　based　on　temporal　clustering　feature.　This　structure　is　suitable　for　extracting　information　from　XML　document　structure　and　calculating　similarity　between　XML　documents.　Secondly,　we　use　the　sliding　window　technique,　which　adopts　exponential　histogram　of　XML　cluster　feature　as　a　micro-cluster　of　it.　By　using　the　model,　we　can　dynamically　accept　the　new　data　and　get　rid　of　the　old　data　thereby　getting　a　better　distribution　feature　of　the　current　window.　Finally,　the　experimental　results　based　on　real　and　synthetic　XML　datasets　show　that　our　algorithm　not　only　achieves　the　real-time　requirements　of　the　online　clustering,　but　also　gains　better　clustering　quality　and　faster　processing　speed.