神经协同过滤智能商业选址方法

A new neural collaborative filtering method, NCF-RS, was proposed based on the framework of neural collaborative filtering (NCF). In order to discover the relationship between store category and site, the linear relationship between store categories and sites was studied by matrix decomposition, the non-linear and in-depth relationship between store categories and sites was studied by deep learning multi-layer perceptron, then the two relationships were combined for the final results. Restaurants' data and POI data in Beijing were used to evaluate the performance of NCF-RS. Results indicate that NCF-RS has better performance in intelligent commercial site recommendation than other advanced deep learning methods and collaborative filtering methods, and can take linear and nonlinear relationships into better consideration.