Interpretable Machine Learning for Complex Human Activity Recognition: A Voting Classifier Approach Using AdaBoost and Random Forest

Human activity recognition is a challenging task due to various factors, such as background clutter, partial occlusion, changes in scale, viewpoint, lighting, and appearance. Moreover, the way that humans perform an activity depends on their habits, making it difficult to identify the underlying activity. This paper presents a novel approach to improve the accuracy of complex activity recognition using accelerometer-embedded mobile phone classifiers. The proposed method employs a voting technique that combines the outputs of multiple classifiers to enhance recognition accuracy while maintaining model interpretability. To evaluate the effectiveness of the proposed approach, a comprehensive evaluation was conducted using publicly available datasets. The results demonstrate that the proposed voting-based ensemble method significantly improved the accuracy of complex activity recognition compared to individual classifiers and deep learning model. The combination of multiple machine learning models, including AdaBoost and Random Forest achieved higher recognition rates for complex activities than any single classifier alone.