Due to the low matching degree and poor recommendation effect of the existing intelligent recommendation methods, we study the cross-border e-commerce multi-collaborative information intelligent recommendation method based on deep learning and graph neural network. Firstly, a heterogeneous user graph is constructed to explicitly model the interaction between users and products. The potential factor vectors of users and products are connected, and then the long vectors obtained after connection are put into the deep neural network for multilayer nonlinear transformation, and finally the predicted values are obtained directly from the output layer. By calculating the similarity, the degree of match between each product and the user's interest is obtained. According to the similarity degree, the candidate products are sorted, and the products that match the user's interest the most are put in the first place, so that they can be noticed and accepted by the user more easily. Finally, the sorted products are selected as the recommendation results. The experimental results show that the matching degree of the experimental group is more than 80%, which is the highest matching degree of the three groups, indicating that the method in this paper can accurately recommend and match the resources, and provide the users with product information recommendation services on the platform.