Fusion recommendation method, system and device

The invention relates to a fusion recommendation method. The method comprises the steps of obtaining behavior data of a plurality of users for a plurality of products; determining the similarity between every two products in the plurality of products based on the behavior data; generating a similar product candidate set of the plurality of products based on the similarity above the threshold; based on the behavior data, calculating scores of at least part of the users of the products in the similar product candidate set to obtain a sparse high-dimensional user-product scoring matrix; performing matrix decomposition on the sparse high-dimensional user-product scoring matrix to mine implicit features of a plurality of users and obtain a decomposed low-dimensional user-implicit feature matrixand a decomposed low-dimensional implicit feature-product scoring matrix; and multiplying the low-dimensional user-implicit feature matrix by the low-dimensional implicit feature-product scoring matrix to obtain a high-dimen