Ethereum Ponzi Scheme Detection Based on PD-SECR

Algorithm 2

CNN-RF detection method.

	input:
	epochs number of training rounds feature, label features and labels of the input data set.
	W processed data set by SMOTEENN
	output:
	The detection result, output the category 0 or 1 of the predicted feature
(1)	W is divided into Train set Wtrain and Test set Wtest and save as PKl file;
(2)	if is_balance = true then load_balance.pkl;
(3)	else load imbalance.pkl;
(4)	setup
(5)	for epoch in range(epochs) do
(6)	for feature, label in Wtraindo
(7)	Training feature extraction model CNN;
(8)
(9)	setup optimizer;
(10)	Evaluate the feature extraction model CNN;
(11)	save best feature extraction model as best.pt and return it;
(12)	setup scheduler
(13)	Training classification detection model RF;
(14)	if is_balance = true then load_balance.pkl;
(15)	else load imbalance.pkl;
(16)	load(best.pt);
(17)
(18)	Classification model for classification detection;
(19)
(20)	ifthen return 1;
(21)	else otherwise return 0;