Approximations based on random Fourier features have recently emerged as an efficient and elegant methodology for designing large-scale machine learning task.Unlike approaches used by the Nystr(o)m method which are randomly sampled from the training examples and are therefore data dependent,we take use of random Fourier features,whose basis functions(i.e.,cosine and sine functions)are sampled from a distribution independent from the training sample set,to cluster preference data which appear extensively in recommender system.The main idea of our method consists of employing random Fourier features to explicitly represent preference data in feature space.Our explicitly mapping method can significantly speed up eigenvector approximation and benefit prediction speed in preference clustering.The advantage of our proposed two-stage method is that we can save computing time and memory space.Compared with traditional preference clustering,our method solve the problem of insufficient memory and improve the efficiency of the operation greatly.At last,the experimental on movie data sets which containing 100000 ratings,show that the proposed method is effectiveness in clustering accuracy than Nystr(o)m methods and K-means while its speed is faster than these clustering approaches.