String Grammar Unsupervised Possibilistic Fuzzy C-Medians for Gait Pattern Classification in Patients with Neurodegenerative Diseases

Table 7

Comparison of the proposed method with the existing methods.

Method	Classification error rate (%)
ALS versus Healthy (2-class problem)
Our proposed method	96.88±6.25
Symbolic entropy [7]	82
Radial basis function (RBF) neural networks (All-training-all-testing) [8]	93.1
Radial basis function (RBF) neural networks (Leave-one-out) [8]	89.66
Least squares support vector machine (Leave-one-out) [9]	82.8
Radial basis function (RBF) support vector machines [10]	96.79
Meta-classifier [11]	96.1326
HD versus Healthy (2-class problem)
Our proposed method	97.22±5.56
Symbolic entropy [7]	95
Radial basis function (RBF) neural networks (All-training-all-testing) [8]	100
Radial basis function (RBF) neural networks (Leave-one-out) [8]	83.33
Radial basis function (RBF) support vector machines [10]	90.23
Meta-classifier [11]	88.674
PD versus Healthy (2-class problem)
Our proposed method	96.43±7.14
Symbolic entropy [7]	89
Radial basis function (RBF) neural networks (All-training-all-testing) [8]	100
Radial basis function (RBF) neural networks (Leave-one-out) [8]	87.1
Radial basis function (RBF) support vector machines [10]	89.33
Meta-classifier [11]	90.3581
NDDs versus Healthy (4-class problem)
Our proposed method	98.44±3.13
Radial basis function (RBF) neural networks [8]	93.75
K classifier [12]	99.17
DECORATE [12]	94.69
Random Forest [12]	94.69
Radial basis function (RBF) support vector machines [10]	90.63