Table 5.

Comparisons between the present study and some published work.

Reference	Characteristic Features	Classifier	Number of Classified States	Construction Strategy of Training Data Set	CA (%)
Zhang et al. [42]	Divide time series data into segmentations	Deep Neural Networks (DNN)	4	Random selection	94.9
Yao et al. [43]	Modified local linear embedding	K-Nearest Neighbor (KNN)	4	Random selection	100
Saidi et al. [34]	Higher order statistics (HOS) of vibration signals + PCA	SVM-OAA	4	Random selection	96.98
Tiwari et al. [5]	Multi-scale permutation entropy (MPE)	Adaptive neuro fuzzy classifier	4	Random selection +10-fold cross validation	92.5
Zhang et al. [13]	Singular value decomposition	Multi class SVM optimized by inter cluster distance	3	Random selection	98.54
Present work	Weighted permutation entropy of IMFs decomposed by EEMD	SVM ensemble classifier + Decision function	3	Random selection	97.78