With the advancement of economic globalization, the development of cross-border e-commerce is becoming increasingly prosperous. In this paper, we design a cross-border e-commerce product artificial intelligence recommendation model based on convolutional neural network, which integrates the logistics spatio-temporal data and user portrait features to optimize the recommendation effect. After preprocessing the historical user data, clustering analysis is used to construct a multi-dimensional user portrait. The Embedding layer is utilized to process the high-dimensional features of the data, and the convolutional neural network model is trained by combining the MSE loss function. The study shows that the model in this paper gradually improves the recall rate and other three indicators from about 0.4 to about 0.9 in the product recommendation scenarios of Top5, 10, 15 and 20. The time used to complete the recommendation is around 61-64s. The product recommendation accuracy rates are all greater than 0.75.
