Super machine learning: improving accuracy and reducing variance of behaviour classification from accelerometry

Table 2 Parameters for the four models tested

Model	Nbins	Mtry	Ntree	Max depth
Random forest	20, 30, 40 (numeric)	5, 10, 15	200	5, 10, 15
	3 (categorical)

	Nbins	Learn rate	ntree	Max depth	Sample rate
Gradient boosting machine	20, 30, 40 (numeric)	0.1, 0.001	250, 700	5, 10	0.7, 0.8, 0.9
Gradient boosting machine	3 (categorical)	0.1, 0.001	250, 700	5, 10	0.7, 0.8, 0.9

	Lambda	Alpha
Logistic regression and super learner	Range exp(−11) to exp(6)	0–1 by 0.025

Nbins number of bins, Mtry number of splits in branches, Ntree total number of trees grown, Max depth maximum depth to grow the trees (for a detailed description of the model parameters and how they are used see Additional file 3)

ISSN: 2050-3385