Top-K Human Activity Recognition Dataset
Moses Gadebe, Tshwane University of Technology ; Okuthe Kogeda, University of Free State
The availability of Smartphones has increased the possibility of self-monitoring to increase physical activity and behavior change to prevent obesity. However self-monitoring on a Smartphtone comes with some challenges such as unavailability of lightweight classification algorithm, personalized dataset to completely capture bodily postures, subject sensitivity, limited storage and computational power. However, most classification algorithms such as Support Vector Machines, C4.5, Naïve Bayes and K Neighbor relies on larger dataset to accurately predict human activities. In this paper, we present top-k of compressed small personalized dataset to reduce computational cost with increased accuracy. We collected top-k personalized dataset from 13 recruited subjects. After benchmarking our collected dataset we found that the dataset is suitable for tree-oriented algorithm, especially the Random Forest, C4.5 and Boosted tree with accuracy and precision of 100% except for KNN, Support Vector and Naïve Bayes. Further, our top-k personalized dataset improves pruning and overfitting of tree-oriented algorithms. Moreover, the linear consistence of static human activities reveals the potential of our top-k dataset to be replicated to multiple-subject to close subject sensitivity challenge.
Gadebe, M. & Kogeda, O. (2020). Top-K Human Activity Recognition Dataset. International Association of Online Engineering.