Maintenance of balance is provided by using different joint strategies regarding the perturbation conditions. Performance of the joint strategies can be assessed by calculation of their kinematic-based features. The aim of this study was to identify the most effective stability features of the dominant role-playing strategy while standing on an unstable platform. Sixteen healthy young men participated in perturbed standing tests on an unstable platform supported by high and low stiffness springs. Motion capture analysis was employed in the sagittal plane to measure the joints angular rotations using. Path length of angular displacement and velocity, total mean velocity, standard deviation, root mean square and fractal dimension features were extracted for quantitative stability analysis. Then, the K-nearest neighbor (KNN), support vector machine (SVM) and multilayer perceptron neural network (MLP) classifiers in the wrapper feature selection technique were used to classify the low and high stiffness supports of the platform. The results first showed that ankle is the dominant strategy in keeping the balance. All three classifiers revealed acceptable performance for data classification, but by applying the wrapper method and selected velocity-based features, finally the support vector classification with 93.75% accuracy had the highest accuracy and efficiency. This study can provide an early diagnosis of balance problems, standing and joint mechanisms to prevent falls by evaluating the classifiers.