Music Recommendation System and Recommendation Model Based on Convolutional Neural Network

In today’s era of big data with excess information, music is common and everyday, which shows the huge amount data. How to obtain one’s favorite from massive database has become a problem, emergence recommendation systems also inevitable. this paper, we take digital piano as research object, form comprehensive features using spectrum notes, design classification methods convolutional neural networks, further process results algorithms. The basic method algorithm determine structure network model, corresponding training improve parameters on basis typical source model used in system experiment. Historical behavior chooses collect information. Then, it reads audio retrieves Mel, reveals identity music. proposal achieves its goal by denying similarity between customer preferences potential two musical characteristics. Two recommended based networks are tested article. On whole, accuracy user’s feature, higher than rate multicategory user. comparison experiment single-category methods, average user feature 50.35%; features. experimental show that can achieve better results.