Since the locally linear embedding method is sensitive to noise, a new method is presented to solve this problem. This new method, namely robust locally linear embedding method, is constructed by analyzing noise influence on the character of the data set's local neighborhood. Compared with LLE, the RLLE is insensitive to noise and adaptive to the selection of neighborhood, which is verified by theoretical analysis and practical experiments. Therefore, the RLLE can discover the intrinsic structure of the data set and visualize the data better.