In order to effectively deal with the uncertain information in pattern classification, a new evidence K-NN classifier (NEK-NNC) is designed, which combines the Dempster and Yager rules with the nearest neighbor classification. Then, for the problem that there is no training sample and the number of target classes is unknown, an adaptive clustering algorithm base on evidence reasoning (ACAER) is proposed. The initial basic belief assignment and the number of classes of the target are given at random and then updated cyclically by the NEK-NNC until it no longer changes, thus achieving full adaptive clustering of the target data. Several experiments based on the simulation and real data sets are given to test the effectiveness of ACAER with respect to the FCM. The results indicate that ACAER can effectively improve the recognition accuracy.