Abstract: Generalizing from spacial database is an important task in the geographical information system.In the real life,data are accumulated after some time,which makes spacial-attributed generalization on the basis of time-series data have practical meaning.An algorithm is presented.The special time-series data are modeled as curve data.The curves that satisfy the matching measurement are modeled as the same one,which reduces the data of being scanned.Then the generalization results can be acquired by searching for the sub-patterns of each curve and gaining relevant knowledge.