Abstract: Classification is an important data mining task in the past decade Meanwhile, many effective and efficient methods, e g decision tree and Bayes network, have been developed for classifying on large static database However, these methods do not fit to processing over data stream So a new algorithm—CAPE（classification using frequent patterns） is presented to deal with classification over data stream Frequent patterns are imported into classification and used to record data distributing over stream mainly during a certain time slice The experimental results show that the accuracy of classification using frequent patterns over stream is higher in most cases compared with the algorithm “weighted classifier ensembles” which is known to be the best classification algorithm over stream at present