Human Activity Recognition and Classification using Ant Colony Optimization based Machine Learning Algorithms

In the recent past, recognizing and tracking uncharacteristic behavior in elderly people is essential due to ageing. According to the survey of World Health Organization (WHO), by 2030, 1 in 6 people will be aged 60 years or over and by 2050, the people aged 60 years and older will 2.1 billion globally. Thus, in this paper, we aim to analyze and assist elderly people in performing daily activities using wearable sensor data. To attain this, an intelligent Human Activity Recognition (HAR) using Ant Colony Optimization (ACO) based Machine Learning algorithms is proposed. It includes two phases namely: 1) Feature Selection and 2) Classification. ACO is used to reduce the redundancy in feature selection from the large data set and helps to upgrade the performance of the classifier in the next stage. Then, we exploit the different classification algorithms such as K-Nearest Neighbors (KNN), Support Vector Machine (SVM), Decision Tree, and Random Forest are utilized for activity recognition using the data collected from wearable sensors. These algorithms are chosen for their ability to deliver high accuracy, computational efficiency, and optimized memory usage, making them well-suited for this application. Furthermore, performance of ACO based machine learning algorithms has been evaluated using accuracy, precision, F1 score, recall and support parameters. The highest classification accuracy, 98.28%, was achieved using the SVM model trained with the ACO-optimized dataset. In comparison, the SVM model without ACO achieved an accuracy of 98.09%, along with a Precision of 98.02%, Recall of 93.22%, and F1-Score of 96.52%. These findings underscore the effectiveness of combining ACO with machine learning techniques for HAR analysis.