Semi-supervised adaptive method for human activities recognition (HAR)

Using sensors and mobile devices integrated with hardware and software tools for Human Recognition Activities (HAR), is a growing scientific field, the analysis based on this information have promising benefits to detect regular and irregular behaviors in individuals during their daily activities. In this study, the Van Kasteren dataset was used for the experimental stage, and it all data was processed using the data mining classification methods: Decision Trees (DT), Support Vector Machines (SVM) and Naïve Bayes (NB). These methods were applied during the training and validation processes with the proposed methodology, and the results obtained showed that all these three methods were successful to identify the cluster associated to the activities contained in the Van Kasteren dataset. The Support Vector Machines (SVM) method showed the best results with the evaluation metrics: True Positive Rate (TPR) 99.2%, False Positive Rate (FPR) 0.6%, precision (99.2%), coverage (99.2%) and F-Measure (98.8%).

• Autores: Mendoza Palechor, Fabio; Vicario, Enrico; PATARA, FULVIO; De la Hoz Manotas, Alexis Kevin; Molina Estren, Diego
• Tipo de recurso: Part of book
• Fecha de publicación: 2022
• Institución: Corporación Universidad de la Costa
• Repositorio: REDICUC - Repositorio CUC
• Idioma: eng
• OAI Identifier: oai:repositorio.cuc.edu.co:11323/9522
• Acceso en línea: https://hdl.handle.net/11323/9522; https://doi.org/10.1007/978-3-031-10539-5_1; https://repositorio.cuc.edu.co/
• Palabra clave: HAR; Data mining; Cluster; Evaluation metric; Dataset; Van Karesten
• Rights: embargoedAccess
• License: © 2022 The Author(s), under exclusive license to Springer Nature Switzerland AG