Comparative Analysis of the K-Nearest Neighbor Method and Support Vector Machine in Human Fall Detection

Falling events occur every day, especially among elderly patients or patients in need of special care. To some extent, falling can cause serious injuries, both open and internal wounds. Therefore, fast and accurate technology is needed to detect whether someone has fallen, especially older adults. This research aims to find a solution for detecting fall movement using machine learning algorithms. In this research, the dataset was collected from 206 participants through the Phyphox application on smartphones. This app utilizes the accelerometer sensor to measure acceleration on each 3D axis (x, y, and z). Here, two proposed classification methods are used to detect the fall movement, i.e., K-Nearest Neighbor and Support Vector Machine. The accuracy of both models is calculated and compared in this paper. The research results show that the linear kernel Support Vector Machine (SVM) has a performance range of 77% to 95%. Meanwhile, the RBF kernel SVM demonstrates a performance range of 79% to 87%, and the polynomial order three of the SVM kernel displays a performance range of 51% to 87%. Moreover, the performance of the K-NN algorithm falls within the scope of 72% to 84%. Overall, the scenario employing the linear kernel Support Vector Machine (SVM) technique exhibits the highest level of efficacy, boasting an accuracy rate of up to 95%.