Based on the stream data with the characters such as real-time, continuous, orderly and unlimited, the continuoustime
data sequence can be detected by using the approximate method. Based on this, making use of samples not only from the
target distribution but also from similar distributions, Tr-OEM algorithm is proposed to detect the concept drift phenomenon
in stream data. This algorithm dynamically estimates the occurrence of concept drift in stream data, automatically determines
optimizing or reconstructing classifiers, and is applied to different types of stream data. The analysis and simulation
experiments show that the proposed algorithm has better adaptability while handling the concept drift in stream data.