The project aims to create a system for human activity recognition using data from smartphone sensors. We have to develop a more efficient system using machine learning algorithms for more accurate activity classification and to optimize the system. HAR has several applications in fitness tracking and sports performance analysis used on smartphones. The six daily actions are walking, walking up and down stairs, sitting, standing, and laying. The application’s main goal is to identify users’ movements using built-in sensors, such as those that monitor acceleration. The project make practical feature selection for the univariate filter method and feature extraction with linear discriminant analysis (LDA). It evaluate multiple models, including the Decision tree, Random Forest, and Support Vector Classifier, using confusion metrics like accuracy, recall, f1score, and precision. The combination of feature selection and extraction exchanges model performances. Additionally, the comparison of the ensemble models’ total accuracy for the classification rate is used to gauge performance. In general, a voting classifier based on a random ensemble model fared worse than a decision tree.