Energy Informatics

Table 3 In each run, a random combination of hyperparameters is tested

From: Load forecasting for energy communities: a novel LSTM-XGBoost hybrid model based on smart meter data

Hyperparameter	Space	Distribution
N estimators	[40,1000]	Randint
Max depth	[1, 100]	Randint
Learning rate	[0.01, 0.59]	Uniform
Subsample	[0.3, 0.6]	Uniform
Colsample bytree	[0.5, 0.4]	Uniform
Min child weight	[0.05, 0.1, 0.02,1, 2, 3, 4]	None
Gamma	[0,0.5,2,10]	None

The number of nestimators and the maxdepth is determined by drawing random integers. Minchildweight and gamma are determined by drawing from a set of pre-determined values. The remaining hyperparameters are drawn from uniform distributions

Back to article page