Empirical Evaluation of Noise Influence on Supervised Machine Learning Algorithms Using Intrusion Detection Datasets

Table 11

Used parameters values for each ML algorithm in WEKA.

ML algorithm	Parameter name	Value
DT	Batch size	100
	Binary splits	False
	Collapse tree	True
	Confidence factor	0.25
	Debug	False
	Do not check capabilities	False
	Do not make split point actual value	False
	Min num obj	2
	Num decimal places	2
	Num folds	3
	Reduced error pruning	False
	Save instance data	False
	Seeds	1
	Subtree raising	True
	Unpruned	False
	Use Laplace	False
	Use MDL correction	True
SVM	Batch size	100
	Build calibration models	False
	C	1
	Calibrator	Default
	Checks turned off	False
	Debug	False
	Do not check capabilities	False
	Epsilon	1.0E − 12
	Filter type	Normalize training data
	Kernel	Default
	Num decimal places	2
	Num folds	−1
	Random seed	1
	Tolerance parameter	0.001
NB	Batch size	100
	Debug	False
	Display model in old format	False
	Do not check capabilities	False
	Num decimal places	2
	Use kernel estimator	False
	Use supervised discretization	False
RF	Bag size percent	100
	Batch size	100
	Break ties randomly	False
	Calc out of bag	False
	Compute attribute importance	False
	Debug	False
	Do not check capabilities	False
	Max depth	0
	Num decimal places	2
	Num execution slots	1
	Num features	0
	Num iterations	100
	Output out of bag complexity statistics	False
	Print classifiers	False
	Seed	1
	Store out of bag predictions	False
ANN	GUI	False
	Auto build	True
	Batch size	100
	Debug	False
	Decay	False
	Do not check capabilities	False
	Hidden layers	A
	Learning rate	0.3
	Momentum	0.2
	Nominal to binary filter	True
	Normalize attributes	True
	Normalize numeric class	True
	Num decimal places	2
	Reset	True
	Seed	0
	Training time	500
	Validation set size	0
	Validation threshold	20