In the fast-paced world of e-commerce, personalized recommendation systems are crucial for enhancing user experience and driving sales. Traditional models such as collaborative filtering (CF) and content-based filtering face challenges like data sparsity, scalability issues, and limited real-time processing capabilities. This study explores the application of graph neural networks (GNNs) to overcome these limitations and improve real-time personalization in e-commerce recommendation systems. GNNs are particularly effective in modeling complex relationships within graph-structured data, allowing for a nuanced understanding of user-item interactions. Using the RetailRocket Recommender System Dataset, this research investigates the effectiveness of GNNs in delivering more accurate and timely recommendations compared to traditional methods. The study’s objectives include assessing the performance of GNNs in capturing complex user-item interactions, evaluating their scalability and real-time processing capabilities, and comparing their performance with other recommendation models. The methodology involves data collection and preprocessing, implementing a GNN-based recommendation model, and integrating this model with a real-time data processing framework. Key performance metrics such as precision, recall, F1-score, and real-time processing latency are used to evaluate the models. Experimental results show that GNNs outperform baseline models in terms of precision, recall, and overall recommendation quality, achieving an average precision of 0.82, recall of 0.79, and AUC of 0.85. The GNN model also demonstrates acceptable real-time processing latency, with an average latency of 120 ms. These findings highlight the potential of GNNs to significantly enhance the effectiveness and efficiency of recommendation systems in e-commerce by providing personalized and timely suggestions. The study concludes that GNNs represent a promising direction for future research and development in personalized recommendation systems, offering a robust solution to existing challenges in the field.