Multi-source heterogeneous data hybrid recommendation model based on deep learning

In recent years, deep learning is widely applied to the fields of image and audio recognition, text classification, representation learning and the like, and a recommendation system based on deep learning also becomes a research hotspot of the scholars. A deep learning model obtains an extremely good effect in the representation learning of the specific data, such as images, texts, etc., a complex feature engineering is avoided, the nonlinear and multi-level abstract feature representation of the heterogeneous data can be obtained, and the heterogeneity of multiple kinds of data is overcome. At present, a deep learning recommendation model fusing scores, comments and social networks is not put forward yet. Based on a deep learning algorithm, a recommendation process with higher expansibility is given, the related algorithms and principles suitable for different data are analyzed, a final loss function combined with comments, scores and social information is derived according to the loss functions of the diffe