Abstract: Selection of parameters wz,wμ,εplays an important role in rough C; Means algorithm In this paper, a dynamic threshold rough C-Means algorithm was proposed to self-adaptive adjusting threshold。that reflects the superposition between classes. This algorithm computes a threshold for every object on the basis of class interval and the distance between class and o场ect, The better effect can be testified by two synthetic data and image data experiments.

Key words: C-Means clustering, Rough sets, Rough C-Means clustering