Estimating Children Engagement Interacting with Robots in Special Education Using Machine Learning

Table 3

Settings of the ML models.

ML model	Best parameters
SVM (RBF)	C = 10, tol = 0.1, and gamma = 0.001
SVM (poly)	C = 10, tol = 0.0001, and gamma = “scale”
Decision Tree	Splitter = “best,” min samples leaf = 10, criterion = “gini,” max features = none, and max depth = 4
k-NN	n neighbors = 2, weights = “uniform,” leaf size = 20, and algorithm = “ball tree”
Naïve Bayes	(Nothing to configure)
MLP	Solver = “Adam,” learning rate = “constant,” hidden layer sizes = (80, 40), tol = 10.0, and alpha = 0.01
ELM	Alpha = 100, n_hidden = 80, and rbf_width = 0.256
Random Forest	Max features = “sqrt,” n estimators = 4, criterion = “gini,” max depth = 15, and min samples leaf = 15
AdaBoost Decision Tree	Criterion = “entropy”, max depth = 15, max features = “auto,” splitter = “best,” and min samples leaf = 5
AdaBoost Naïve Bayes	(Nothing to configure)