Table 3 The overall comparison between linear, MLP, and deep RNN models.

From: Development of a hip osteoarthritis index for gait quality assessment: a data-driven comparative study

Mean±SD

Model accuracy (%)

Linear model

MLP model

Deep RNN model

Healthy-patient

84 ± 8

89 ± 5

89 ± 4

Affected-unaffected leg

91 ± 11

90 ± 7

88 ± 5

Severity grades

41 ± 9

42 ± 12

39 ± 9

  1. The comparison between the classification accuracy of three different data-driven models in three different scenarios: (1) healthy-patient, (2) affected-unaffected leg, and (3) severity grades classifications. The linear model results correspond to using the best features (F6 and F10) as input, the MLP results represent the entire data with general risk factors condition, and the deep RNN results relate to the perfect unseen with general risk factors condition. The paired comparisons between the models’ classification rates show no statistically significant differences; \(p>0.16\) for all paired comparisons.
Back to article page