Impact of feature reduction techniques on classification accuracy of machine learning techniques in leg rehabilitation

Abstract

The scientific community has recently shown a lot of interest in biometric-based person recognition systems. It is a dynamic technology that tries to recognize biometrics automatically, swiftly, precisely, and consistently. Gait recognition is a sort of biometric classification that concentrates on identifying persons utilizing personal measurements and correlations, including trunk and limb size, in addition to spatial data connected to innate patterns in people's motions. The purpose of this study is to highlight how important it is to use feature reduction strategies to increase classification accuracy. Two of these approaches are employed in this work: Principal Component Analysis (PCA) and Singular Value Decomposition (SVD). To classify foot disease, six machine learning techniques are deployed as classifiers. These classifiers are Random Forest (RF), Naïve Bayes (NB), K-Nearest Neighbors (KNN), Logistic Regression (LR), Decision Tree (DT), and Stochastic Gradient Descent (SGD) for the purpose of determining which classifier performs best in classifying leg rehabilitation data. Experimental results using the EMG dataset in Lower Limb indicate that the classification accuracy reached 99% with a time not exceeding a few seconds.