Abstract: A method based on Adaptive Hierarchy Aggregation tree(AHA-Tree) was presented for computing aggregalion of data stream. The tructure of AHA-Tree borrowed the idea of multiple time granularities hierarchical window model,the recent data was kept in fine granularity and the older in rough. In addition,the partition of granularities was　determined by density of time unit, the sparse time unit was kept in rough granularity and the denseness in fine. Moreover, the method of maintenance and aggregate computing was proposed for aggregation query. Experiment shows that the method is efficient in processing the data under non-uniform distribution.

Key words: Data stream, Aggregate computation, Adaptive hierarchy aggregation tree, Time window