A Novel Dynamic Chaotic Golden Jackal Optimization Algorithm for Sensor-Based Human Activity Recognition Using Smartphones for Sustainable Smart Cities

As to the World Health Organization’s Sustainable Development Agenda 2030, engaging in regular physical activity offers numerous societal benefits for creating healthier communities and society. The integration of the Internet of Things (IoT) and widespread smartphones is crucial for achieving a major advancement in different areas of smart cities, such as healthcare, fitness, skill evaluation, and personal assistants, to facilitate independent living. The devices supported by the Internet of Things (IoT) are equipped with sensors that allow various context-aware applications to identify and understand physical activity. While there are activity recognition applications, their ability to reliably identify activities is currently lacking. The effectiveness of HAR systems heavily depends on the selection of relevant features to accurately classify and recognize human activities. This paper proposes an improved variant of Golden Jackal Algorithm named IGJO. Two improvements are integrated into GJO including the dynamic opposite learning and chaotic maps to enhance the exploration phase and the global search process of the algorithm. The proposed IGJO is applied for feature selection in HAR to improve the classification accuracy. The proposed IGJO was tested using global optimization functions besides the feature selection. The experimental results proved that IGJO obtained an accuracy of 96.69% and recall of 96.84% compared with other five related top contributed algorithms. The experimental results proved the superiority of the proposed IGJO in optimization accuracy and convergence speed.