Learn to construct a Graph Convolutional Network that may handle heterogeneous graph data for link prediction
This text is an in depth technical deep dive into learn how to construct a robust model for anomaly detection with graph data containing entities of differing types (heterogeneous graph data).
The model you’ll find out about is predicated on the paper titled “Interaction-Focused Anomaly Detection on Bipartite Node-and-Edge-Attributed Graphs” presented by Grab, an Asian tech company, on the 2023 International Joint Conference on Neural Networks (IJCNN) conference.
This Graph Convolutional Network (GCN) model can handle heterogeneous graph data, meaning that nodes and edges are of differing types. These graphs are structurally complex as they represent relationships between various kinds of entities or nodes.
GCNs that may handle heterogeneous graph data is an energetic area of research. The convolutional operations within the model have been adapted to deal with challenges around handling different node types and their relationships in a heterogeneous graph.
In contrast, homogeneous graphs involve nodes and edges of the identical type. The sort of graph is structurally simpler. An example of a homogeneous graph include LinkedIn connections, where all nodes represent individuals and edges exist between individuals in the event that they are connected.
The instance you will notice here applies Grab’s GraphBEAN model (Bipartite Node-and-Edge-Attributed Networks) to a Kaggle dataset on healthcare provider fraud. (This dataset is currently licensed CC0: Public Domain on Kaggle. Please note that this dataset won’t be accurate, and it’s utilized in this text just for demonstration purposes). The dataset incorporates multiple csv files with claims and insights on inpatient data, outpatient data, and beneficiary data.
I’ll reveal learn how to construct a GCN to predict healthcare provider fraud using the inpatient dataset and train set containing ProviderIDand a label column (PotentialFraud).
While graph data will be difficult to visualise in tabular form, just like the csv files, you possibly can make interesting…