AMRM: machine learning based intelligent recommendation of preference of music

With an increasing quantity of music applications in recent years, recommendation of music became the key determination for these apps to attract users. However, for sun-rise music app firms, they are not able to obtain enormous data of users to create their AI recommendation mechanism to catch the music preference of users accurately. In this paper, we introduce an artificial intelligent recommendation model called AMRM, which can analyse a small amount of dataset related to end users to recommend the songs which match the music tastes of the users to the best extent, for small companies to create their recommendation mechanism. Our method is proposed based on the real world questionnaires about the features in user dimension and the information learning models. The traditional machine learning model of GBDT and the deep-learning model CNN are applied to compare one other to find out which are the users' favorite songs on the Netease Cloud. Based on the evaluation experiments, the experimental results show that the traditional machine learning model GBDT is the more feasible and more appropriate to the analysis on our small-sized dataset.