“…The learning algorithms mainly use raw data or extracted features along with classification methods, such as artificial neural networks (ANN) [ 12 , 13 , 16 , 20 , 21 , 24 ], random forests (RF) [ 14 , 15 , 19 , 23 , 26 ], support vector machines (SVM) [ 22 , 23 ], and k -nearest neighbours (KNN) [ 26 ], among others. For example, in [ 20 , 21 ], restricted Boltzmann machines are trained using features extracted from wrist-worn accelerometer data in a home environment to predict PD state.…”