Unveiling Novel Techniques for Human Activity Recognition with Smartphone Sensors

Due to the accessibility and capability of smartphones, human activity recognition (HAR) has attracted a lot of interest. HAR includes using sensor data gathered from cellphones to automatically recognise and categorise human behaviours. Labelled sensor data gathered from cellphones made up the benchmark dataset utilised in this investigation. A wide variety of human motions and activities were included in the dataset, which also included various postures and movements.In this work, we assess the performance of five well-known classification techniques for HAR: k-nearest Neighbours (KNN), Linear Support Vector Machines (SVM), Random Forest, Decision Trees, and Logistic Regression.We use a benchmark dataset made up of labelled sensor data gathered from cell phones to assess the algorithms. The dataset includes a variety of human motions, including sitting, standing, jogging, and other bodily actions.The early findings show that the algorithms' degrees of accuracy differ. With a high accuracy score of 0.954869, logistic regression demonstrates great predicting ability. With an accuracy score of 0.872413, Decision Trees perform slightly worse than Random Forest, which performs better overall with a score of 0.921955. The most effective method, with an outstanding accuracy score of 0.968442, is linear SVM. The accuracy score for k-Nearest Neighbours, which is 0.902273, is also good. This side-by-side comparison sheds important light on how various HAR algorithms for cell phones function. The results support the creation of more precise and effective HAR systems by assisting researchers and practitioners in choosing the best algorithm for their particular applications. The next sections of this study give further information on the methodology, experimental design, findings, and commentary.