I am working on anomaly detection and have a dataset with multiple graphs. For instance, one graph has 100K edges and it belongs to video game activities (API calls). Another graph belongs to attack and so on. Now, I would like to create some baseline datasets with multiple activities. For instance, a dataset that contains some video game activities and some download activities and some attacks, etc. How can I do that accurately? It seems I should consider statistical measures for a good dataset. You can see the dataset here. I would like to create YDC for example.