Experimental results show that this method is of low memory requirements and lower mean absolute error (MAE) value, and provides better recommendation quality compared with traditional collaborative filtering algorithms.