A Hyperparameter Optimization Algorithm for the LSTM Temperature Prediction Model in Data Center

Table 3

Algorithm model and corresponding hyperparameters.

Algorithm	Hyperparameter	Ranges	Interval
LSTM	Act.ivation	[“linear,” “sigmoid”, “relu,” “tanh”]	0.1
	Dropout	[0∼0.5]	10
	Unit	[20∼200]	N ^∗2
	Batch_Size	[32, 64∼512]
	Optimizer	[“SGD,” “adagrad,” “Adadelta,” “Adam,” “nAdam”]
Random forest	n-estimators max_depth	[100∼1200]	100
	min_samples_split	[2∼30]	2
	min_samples_leaf	[1∼99]	2
	max_features	[1∼9]
	criterion	[sqrt, log2, None]
	bootstrap	[gini, entropy] [True, False]	2