Research on cross-border e-commerce product marketing recommendations based on multimodal signal fusion and deep learning technologies

With the rapid advancement of artificial intelligence and big data technologies, cross-border e-commerce platforms have experienced global expansion. Traditional product recommendation methods demonstrate limited effectiveness in cross-domain and multimodal scenarios, failing to fully leverage deep correlations between multi-source information such as images, text, and videos. This limitation has constrained both the precision of recommendation systems and user experience. To address these challenges, this study proposes a multimodal intelligent recommendation model integrating graph convolutional networks and Transformers, designed to enhance the accuracy, diversity, and real-time performance of product recommendations in cross-border e-commerce environments. Experimental results show that the proposed model achieves 39.2% higher precision, 39.2% better recall rate, and 94% higher recommendation match rate compared to traditional multimodal models on real cross-border e-commerce datasets, with a response time of merely 23 milliseconds. This research not only provides an efficient and scalable multimodal integration solution for cross-border e-commerce recommendation systems but also significantly improves recommendation quality and user experience in practical applications. The findings hold substantial theoretical and practical value for advancing intelligent recommendation technologies in global e-commerce ecosystems.