Hierarchical Summarization of Homogeneous Records Using Graph Neural Networks

In today’s data-driven world, the inefficient and inconsistent recalculation of hierarchical summaries presents a significant challenge for organizations relying on large datasets for critical decision-making. Traditional methods often suffer from delays and data inconsistencies as datasets expand. To address this issue, a novel solution using Graph Neural Networks (GNNs) is proposed for asynchronously recalculating hierarchical summaries in uniform records. This approach enhances data accuracy, accelerates processing, and maintains scalability by transforming hierarchical datasets into graph structures, allowing GNNs to capture complex relationships. Additionally, a specialized batching method is introduced to allow for simultaneous updates without locking entire datasets. Extensive testing demonstrates that this method significantly outperforms traditional recalculation techniques in terms of speed, consistency, and scalability, providing a robust and scalable solution for managing large, complex data structures in real-world applications