Song recommending method based on nonlinear manifold learning

The invention belongs to the field of computer technique recommending methods, and relates to a song recommending method based on nonlinear manifold learning. The song recommending method mainly comprises the following steps of establishing a relationship model between a user and a song; performing classifying and dimensional reduction on the song; according to the relationship model obtained after dimensional reduction, recommending by the song recommending method. The song recommending method has the beneficial effect that compared with the traditional synergistic filtering algorithm, more accuracy is obtained, and a recommending plan is more reasonable. 本发明属于计算机技术推荐方法领域，涉及种基于非线性流形学习的歌曲推荐方法。本发明的主要方法包括：建立用户与歌曲的关系模型；对歌曲进行分类降维操作；根据降维后获得的关系模型再通过歌曲推荐方法进行推荐。本发明的有益效果为，推荐较之传统协同过滤算法更加准确，推荐方案更加合理。