Edge Classification

[PrashnikB]

I am trying to make a GNN model for credit card fraud detection.
The graph that I built has users as nodes and the transactions as edges I want to label each edge as fraud or not but I am not able to find any example where edge label has been been predicted. I am not trying to do link prediction, I aldready have the edges I only want to classify them as fraud or not fraud.
Can someone help

[xnuohz]

https://github.com/pyg-team/pytorch_geometric/blob/master/examples/link_pred.py

[PrashnikB]

In this example they are doing link prediction but I aldready have the links I want to do edge classification.

[PrashnikB]

Is there anyway to use link prediction for edge classification?

[xnuohz]

IMO, link prediction is similar to edge classification.
In edge classification, you could first use GNN extract node and edge features, then define a downstream function like f(node_src_feature, edge_feature, node_dst_feature) to predict the edge's label.