Integrating Feature Engineering with Deep Learning to Conduct Diagnostic and Predictive Analytics for Turbofan Engines

Table 2

Brief comparison in machine learning and deep learning.

Machine learning	Main adjustable hyper-parameters
MARS	Degree of nonlinearity (number of knots), cost penalty
RF, XGB	Minimal samples to split a node, maximal depth, number of trees, learning rate, split ratio between training and testing
SVM	Cost penalty, spread of Gaussian kernels (gamma), width of tolerance (epsilon)
CNN, DNN, RNN, LSTM, GRU, CNN	Number of hidden layers, number of neurons, activation functions, optimizers, drop-out rates, learning rates, dense layers, initializers