Electricity Theft Detection in Power Grids with Deep Learning and Random Forests

Table 4

Experimental parameters for comparative methods.

Methods	Input data	Parameters
LR	Features (1D)	Penalty: L2
		Inverse of regulation strength: 1.0
SVM	Features (1D)	Penalty parameter of the error term:
		Parameter of kernel function (RBF): 0.0005
GBDT	Features (1D)	The number of estimators: 200
RF	Features (1D)	The number of trees in the forest: 100
CNN	Raw data (2D)	The max depth of each tree: 30
		The same as the CNN component of the proposed approach