The invention discloses a method and device for recommending music. The method is applied to a music recommending device and comprises the following steps of: S1, acquiring a music membership function which is the aggregation of granulations of music in different dimensionalities, wherein the dimensionality is the classification of music and the granulation is the classification of the dimensionalities; S2, acquiring a user membership function which is favored by users and is the aggregation of granulations in different dimensionalities, S3, calculating a granularity relevant function according to the music membership function and the user membership function; S4, calculating a value of a probability function of music favored by the users through the granularity relevant function and a dimensionality weighting coefficient; and S5, recommending the music to the users when the probability value of the music favored by the user is greater than the preset threshold. The invention can provide the individual recommendation for the users so as to improve the degree of satisfaction of the users.