Table 2.

Comparison of the results of training and testing of classification models for two and six levels of fall risk.

Model	2 Fall Risk Levels (i.e., Fallers and Nonfallers)					6 Fall Risk Levels (i.e., Very Low, Low, Mild, Moderate, High, and Very High)
	acc_Train	10fcv	acc Test	G_Mean	F1-Score	acc_Train	10fcv	acc Test	G_Mean	F1-Score
Logistic Regression	0.838	0.786	0.692	0.673	0.69	0.623	0.409	0.41	0.617	0.41
Ridge Classifier	0.818	0.753	0.692	0.663	0.69	0.5	0.364	0.333	0.482	0.33
LASSO	0.838	0.773	0.718	0.704	0.72	0.565	0.364	0.333	0.508	0.33
K-nearest Neighbours	0.838	0.721	0.641	0.629	0.64	0.623	0.364	0.333	0.535	0.33
Naive Bayes	0.76	0.753	0.692	0.682	0.69	0.468	0.331	0.385	0.602	0.38
Linear Discriminant Analysis	0.818	0.753	0.692	0.663	0.69	0.597	0.396	0.436	0.626	0.44
Decision Tree	1	0.643	0.667	0.67	0.67	1	0.331	0.436	0.631	0.44
Perceptron	0.76	0.746	0.667	0.651	0.67	0.435	0.337	0.282	0.483	0.28
Multilayer Perceptron	1	0.76	0.667	0.651	0.67	1	0.422	0.462	0.657	0.46
Stochastic Gradient Descent	0.805	0.74	0.641	0.629	0.64	0.526	0.382	0.41	0.594	0.41
Gradient Boosting	1	0.772	0.641	0.629	0.64	1	0.434	0.41	0.639	0.41
XGBoost	1	0.721	0.667	0.67	0.67	1	0.363	0.385	0.628	0.38
Support Vector Machine	0.903	0.74	0.615	0.598	0.62	0.662	0.402	0.359	0.491	0.36
Random Forest	1	0.798	0.615	0.598	0.62	1	0.429	0.385	0.584	0.38
AdaBoost	1	0.772	0.564	0.564	0.56	0.494	0.389	0.205	0.333	0.21
Multi-head CNN+LSTM	1	0.991	1	1	1	1	0.987	1	1	1

ICC—Intraclass Correlation Index of the three repetitions of the test; MLDisp—range of the medial–lateral displacement of the centre of mass (CoM) in balance phase; APDisp—range of the anterior–posterior displacement of the CoM in balance phase; VRange—range of vertical displacement of the CoM in gait phase; MLRange—range of medial–lateral displacement of the CoM in gait phase; PTurnSit—power of turn to sit; PStand—power to sit to stand; APJerkSit—anterior–posterior jerk to sit; APJerkStand—anterior–posterior jerk to sit and to stand.