Graph Neural Networks for Precision-Guaranteed Compression of Large-scale Data

Recent advancements in information technology and the development of smart infrastructure within smart cities have led to an unprecedented increase in data generation. Such large volumes of data can easily overburden data storage and management systems, potentially impeding the full exploitation of the data. Existing traditional and prediction-based compression methods have been applied to reduce data storage needs. However, these methods often struggle with issues such as limited compression ratios and no direct control over decompression error, particularly when tested on large-scale datasets. To address these challenges, this study, for the first time, explores the application of advanced spatio-temporal models such as Graph Neural Networks (GNNs). Specifically, we propose a novel prediction-based compression method using GNN that guarantees the desired decompression precision. We demonstrate the efficacy of the proposed method on three large-scale datasets, including traffic data, smart meter energy consumption data, and renewable power generation data, each incorporating data from thousands of locations. The results demonstrate the scalability and robustness of the proposed method across different data types. Moreover, it outperforms state-of-the-art methods in the literature in terms of accuracy and compression efficiency.