Hierarchical Graph Convolutional Network for Data Evaluation of Dynamic Graphs

As data are being generated at an incredible speed and scale, the market of data that aims to fully discover the value of data is becoming increasingly important. Data evaluation is a vital part of the data market. It can discover the data’s flaws and the value hidden in the data. Nevertheless, there have been few studies investigating how to evaluate data in the data market. Existing methods seldom utilize the multi-level structure in data. Our work proposes a novel hierarchical graph convolutional network for the data evaluation of dynamic graphs, following the anomaly detection paradigm. Our model performs significantly better than existing models on several benchmark datasets. This study provides a more powerful tool for processing dynamic graphs. It could also guide the direction of data evaluation for a broader range of data categories in the data market.