Table A1. Hyperparameter grid... | Emerald Publishing

Table A1.

Hyperparameter grid

Hyperparameter	Description	Search space	Model
num_iterations	Number of boosting iterations	Fixed to 100 by the developers to Reduce down the search space to Facilitate efficient optimization Learning_rate hyperparameter is Tuned as a proxy for the effect of Lesser or higher number of trees	100
feature_fraction	LightGBM will randomly select A subset of features on each Iteration (tree), if feature fraction is smaller than 1.0. For example, if you set it to 0.8, LightGBM will select 80% Of features before training each tree • can be used to speed up training • can be used to deal with over-fitting	0.3, 0.5, 0.7, 0.9	0.3
max_depth* * Important hyperparameter	Limits the max depth of a tree This is used to deal with overfitting when #data is small. Tree Still grows leaf-wise	2, 4, 8, 10	4
learning_rate* * Important hyperparameter	A technique to slow down the Learning by weighting the Corrections by new trees at Every learning iteration	0.001, 0.01, 0.05, 0.1	0.1
min_gain_to_split	Minimal gain to perform	0.3, 0.5, 0.7	0.3
min_data_in_leaf	Minimal number of data in one Leaf. Can be used to deal with over-fitting	5, 10, 15, 20	15
lambda_l1	L1 regularization	0.1, 0.3	0.3
scale_pos_weight	Weight of labels with positive class	1.0, 1.5, 1.7	1.7

Hyperparameter	Description	Search space	Model
num_iterations	Number of boosting iterations	Fixed to 100 by the developers toReduce down the search space toFacilitate efficient optimizationLearning_rate hyperparameter isTuned as a proxy for the effect ofLesser or higher number of trees	100
feature_fraction	LightGBM will randomly selectA subset of features on eachIteration (tree), iffeature fraction is smaller than 1.0. For example, if you set it to0.8, LightGBM will select 80%Of features before training eachtree• can be used to speed up training• can be used to deal with over-fitting	0.3, 0.5, 0.7, 0.9	0.3
max_depth** Importanthyperparameter	Limits the max depth of a treeThis is used to deal with overfitting when #data is small. TreeStill grows leaf-wise	2, 4, 8, 10	4
learning_rate** Importanthyperparameter	A technique to slow down theLearning by weighting theCorrections by new trees atEvery learning iteration	0.001, 0.01, 0.05, 0.1	0.1
min_gain_to_split	Minimal gain to perform	0.3, 0.5, 0.7	0.3
min_data_in_leaf	Minimal number of data in oneLeaf. Can be used to deal withover-fitting	5, 10, 15, 20	15
lambda_l1	L1 regularization	0.1, 0.3	0.3
scale_pos_weight	Weight of labels with positiveclass	1.0, 1.5, 1.7	1.7

Source(s): Authors’ own creation/work