In this study, Temporal-Aware Heterogeneous Graph Neural Networks (TAH-GNNs), represent a new generation of Cross-Domain Recommendation Systems. In order to address the Cold-Start problem and improve recommendations’ accuracy, we analyze users’ behaviors on two distinct product categories (Beauty Products and Appliances) from the Amazon review datasets. In order to capture changing user preferences and cross-domain knowledge transfer, our model makes use of a dynamic heterogeneous graph structure with temporal attention mechanisms. Results from experiments show that our method outperforms conventional single-domain recommendation systems, with notable gains in hit rate, precision, recall, and NDCG measures. Domain-specific transformations allow for efficient knowledge transfer across the appliance and beauty domains, while the temporal attention mechanism efficiently weights the significance of past encounters according to recency. Significant correlation patterns that may be used to improve recommendations quality are shown by our study on crossdomain user behavior, particularly for users with little prior experience in a single domain.