Machine Learning for Activity Tracking and Multi-inhabitant Recognition in Smart Environments: An Evaluation Study

Timothy Musharu (), Dieter Vogts () and Jorge Marx Gómez ()
Additional contact information
Timothy Musharu: Carl Von Ossietzky University of Oldenburg
Dieter Vogts: Nelson Mandela University
Jorge Marx Gómez: Carl Von Ossietzky University of Oldenburg

A chapter in Advancement in Embedded and Mobile Systems, 2026, pp 181-199 from Springer

Abstract: Abstract Smart environments have become increasingly popular due to pervasive sensing technologies that offer unprecedented opportunities for health monitoring, security, fitness tracking, and aiding individuals with difficulties living independently. A primary challenge in these environments is recognising and tracking the functional activities of the inhabitants. This paper explores various machine learning approaches, including Multi-Layer Perceptron (MLP), Deep Neural Network (DNN), Decision Tree (DT), and Random Forest (RF), for recognising inhabitants and tracking their Activities of Daily Living (ADLs) within a smart environment. The methodology involves two main components: user recognition and activity tracking. The process includes data pre-processing, feature extraction and selection, and classifier model construction to distinguish between multiple inhabitants for user recognition. For activity tracking, models process input data and classify the status of activities as started, in progress, or completed. Synthetic data generated from the Smart Environment Simulator (SESIM) with three virtual inhabitants was used for model evaluation. To address the imbalanced nature of the dataset, the Synthetic Minority Over-sampling Technique (SMOTE) was applied to ensure fair model performance. Results indicate that the supervised learning models can recognise inhabitants with up to 92% accuracy, with the RF model achieving the highest accuracy at 92%, outperforming the DT model at 90%. Specifically, the RF model shows significant improvements in classifying inhabitant a1, with precision, recall, and F-score of 92%, 86%, and 89%, respectively. The clustering results demonstrate the effectiveness of grouping activities over time, though further refinement using domain expert knowledge is needed to improve accuracy. These findings contribute both practically and theoretically by demonstrating that machine learning can effectively enhance inhabitant recognition and activity tracking in smart environments, providing a foundation for further development in this field.

Keywords: Smart environments; Activity tracking; Machine learning; Multi-inhabitant recognition (search for similar items in EconPapers)
Date: 2026
References: Add references at CitEc
Citations:

There are no downloads for this item, see the EconPapers FAQ for hints about obtaining it.

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:spr:prochp:978-3-031-99219-3_13

Ordering information: This item can be ordered from
http://www.springer.com/9783031992193

DOI: 10.1007/978-3-031-99219-3_13

Access Statistics for this chapter

More chapters in Progress in IS from Springer
Bibliographic data for series maintained by Sonal Shukla () and Springer Nature Abstracting and Indexing ().