Wayne Huang retweetledi

On-chain fraud detection — I’ve been heads-down on modeling lately, and it’s taken quite a bit of time.
I’m currently experimenting with 8 models. To improve accuracy further, I took one of them and split it into 39 variants for ablation / controlled comparisons.
Dataset: 12,719 nodes (d1) / 23,178 nodes (d2).
So far, it looks like I’ve found a clearly better approach✅
#Web3 #AI #AML

English





















