Fig. 2: Feature selection utilizing LASSO regression and evaluation of feature importance by using random forest methodology.

a, d, g, j illustrate the coefficient path plots for clinical data, gait, eye tracking and gFCD, respectively. b, e, h, k represent binomial deviations for clinical data, gait, eye tracking, and gFCD, respectively. c, f, i, l represent random forest algorithms for clinical data, gait, eye tracking, and gFCD, respectively.