Recent years witness a boom in music recommender systems due to the success of online streaming services. Even though such systems have brought relatively high-quality recommendations to the users, they are still facing the cold-start problem, especially for new user case. This problem happens when the system does not have information about the new user's preferences to provide recommendations. Therefore, effectively predicting musical preferences for the new user becomes vital. In this paper, we leverage a five-factor MUSIC model which is characterized by Mellow, Unpretentious, Sophisticated, Intense, and Contemporary to represent the user's preference. Then, towards solving the new user cold-start problems in the content-based music recommender system, we propose a method to predict the five-factor preference profile of the novel user. We consider an early-stage scenario when there are no and few rating data of the user available in the system. Accordingly, we first use the information of age and brain type extracted from questionnaires to build regression models. These models are used to predict the first five-factor musical preference profile for initial recommendations. We then estimate the second five-factor profile based on the user's rating data and linearly combine it with the first profile for improving recommendations. The results demonstrated the effectiveness of the proposed method in predicting the musical preference of the new user in the assumed scenario.