Monte Carlo Schemata Searching for Physical Activity Recognition

摘要: Medical literature have recognized physical activity as a key factor for healthy life due to its remarkable benefits. However, there is great variety of activities and not all them the same effects on health nor require effort. As result, ubiquity commodity devices able track users' motion, an increasing interest performing recognition in order detect type carried out by subjects being credit their effort, which has been detected requirement promote activity. This paper proposes novel approach using Monte Carlo Schemata Search (MCSS) feature selection random forests classification. To validate this we evaluation over PAMAP2, public dataset available UCI Machine Learning repository, enabling replication assessment. The experiments are conducted leave-one-subject-out cross validation attain classification accuracies 93% roughly one third total set features. Results promising, they outperform those obtained other works significantly reduce features used, could translate decrease number sensors required perform and, reduction costs.