YAŞA BAĞLI DUYUSAL DEĞİŞİKLİKLERİN POSTURAL SALINIM DÜZENSİZLİĞİ ÜZERİNDEKİ ETKİSİNİN DEĞERLENDİRİLMESİ

Age-related decline in sensory inputs in elderly people leads to postural instability that increases irregularity of postural sway. This study aimed to examine the effect of visual or somatosensory inputs on postural sway irregularity in the elderly by using machine learning (ML). The feature set was extracted from entropy measurements including sample, fuzzy, distribution, conditional, and permutation. Then, the variables were classified by ML including support vector machines (SVM), k-nearest neighbors (k-NN), and linear discriminant analysis (LDA) algorithms. Classification performances were compared with the confusion matrix. For the elderly, in the eyes closed condition on an unstable surface, the SVM algorithm achieved higher accuracy (77%), sensitivity (72%), specificity (85%), and precision (83%) for the cv dataset. For young, SVM also achieved high accuracy (86%), sensitivity (87%), specificity (84%), and precision (84%). For the elderly, under the eyes open on unstable surface conditions, the SVM exhibited an accuracy of 79%, sensitivity of 75%, specificity of 72%, and precision of 75%. However, for young, it did not reveal good results for both surfaces. In conclusion, the findings suggest that older people adapt their postural control mechanisms, relying more on somatosensory inputs. ML algorithms with entropy-based features can give insights into age-related differences in postural control.