Today, humans can recognize their daily activities by using smartphone activity recognition. Numerous studies have been conducted to identify activities, but for some reason the performance of classifiers is poor due to various issues with the data or the classifiers themselves. This study provides a way to obtain the most effective classifiers.For this reason, a comparative approach was used to assess the effectiveness of supervised and Consequently, ensemble learning classifiers. In this study, the authors also present a system which is based on the best per- forming of the classifier mentioned above. Two publicly accessible datasets of recognized human activities from the UCI repository are used to assess the approach. The first is Human Activities based recognition and Postural Transitions, and the second is UCI-Human Activity Recognition. For this research study, the many different activities have been chosen like walking, standing, sitting, lying down, and upstairs. HAR entails the recognition and categorization of Human Activities from information gathered from sensors, cameras and other wearable gadgets. It has become increasingly essential in healthcare for monitoring and early detection, in fitness and sports for performance tracking, and in security for suspicious behavior identification.