Digital commerce intelligent cross-domain recommendation method integrated with user portrait and social relation similarity

The invention discloses a digital commerce intelligent cross-domain recommendation method integrated with user portrait and social relation similarity. The method comprises the following steps of: firstly, constructing a user portrait vector by adopting a 0-1 scalar and integrating age, gender, project preference and the like of a user, and calculating the similarity of user portraits according to the user portrait vector; secondly, according to the method, improving a common neighbor algorithm, and fusing first-order and second-order neighbors to calculate the user social relation similarity; on the basis, fusing user portrait similarity and user social relation similarity, decomposing a scoring matrix, representing related parameters by using a probability graph model, subjecting an obtained objective function to parameter learning by using an alternating least square method, and finally, using the optimized parameters for scoring prediction. In the whole recommendation process, basic information, preferenc