Compact target-aware stability-based feature selection for cross-dataset wearable activity monitoring
- 1Department of Information Systems, College of Computer and Information Sciences, Princess Nourah bint Abdulrahman University, Riyadh 11671, Saudi Arabia
- 2LR11TIC04, Communication Networks and Security Research Lab, and LR11TIC02, Green and Smart Communication Systems Research Lab, Higher School of Communication of Tunis (SUP’COM), University of Carthage, Ariana, Tunisia
- 3College of Applied Computer Science, King Saud University, Riyadh 11451, Saudi Arabia
Abstract
Wearable activity-recognition models may lose reliability when transferred across participants, sensor placements, sampling rates, acquisition protocols, and activity definitions. This study evaluates a compact unsupervised target-aware feature-selection strategy for cross-dataset wearable activity monitoring under a pre-specified MHEALTH→PAMAP2 transfer protocol. Five-second wearable-signal windows were represented using 270 handcrafted inertial and magnetic descriptors. Source-domain mutual-information relevance was then combined with unlabeled source–target descriptor stability to select a compact set of 48 features, after which lightweight classifiers were trained using MHEALTH labels only. The original activities were harmonized into three coarse states: sedentary, active, and vigorous, while harmonized PAMAP2 target labels were withheld from model development and used only for final evaluation. Under leave-one-subject-out evaluation on MHEALTH, the Random Forest classifier achieved 0.9985 accuracy and 0.9984 macro-F1. In the primary cross-dataset experiment, the pre-specified stability-aware configuration achieved 0.9042 accuracy and 0.8753 macro-F1 on PAMAP2, compared with 0.8888 accuracy and 0.8436 macro-F1 for the full 270-feature representation and 0.8734 accuracy and 0.7940 macro-F1 for source-mutual-information-only selection. With all operating values frozen, reverse PAMAP2→MHEALTH transfer also ranked Stable-48 highest (accuracy 0.5811; macro-F1 0.4400), ahead of Source-MI-48 (0.5720; 0.4320) and Full-270 (0.5445; 0.4257), although Holm-adjusted participant-level contrasts were not significant. The compact configuration also achieved higher complete-window accuracy and macro-F1 than the tested shallow target-feature transformations and feature-level MLP adaptation baselines within the same handcrafted descriptor space. These findings indicate that jointly considering source-domain relevance and unlabeled cross-domain descriptor stability can improve compact cross-dataset wearable activity monitoring. The conclusions remain specific to the evaluated protocols and activity taxonomies and do not establish general superiority over raw-signal deep-learning methods or clinical readiness.
Related
- Development of a Wearable Sensor-PAM for Continuous Monitoring of Photosynthetic Dynamics
- UniHAR: A unified hierarchical transformer with multi-scale temporal modeling for wearable human activity recognition
- Short-term prediction of COPD exacerbations based on wearable vital sign monitoring
- A randomised feasibility trial of VR-enhanced cue exposure therapy versus neutral VR exposure for cocaine use disorder, with wearable monitoring and JITA-EMA: Feasibility randomised controlled trial protocol
- A multimodal deep learning for rice crop health monitoring using vegetation index feature selection
BibTeX
@article{ayouni2026compact,
title = {{Compact target-aware stability-based feature selection for cross-dataset wearable activity monitoring}},
author = {Sarra Ayouni and Maha Sliti and Abeer Abdulaziz Alghanem and Fatimah Alhayan and Mohamed Maddeh},
journal = {PLOS One},
year = {2026},
month = oct,
doi = {10.1371/journal.pone.0359300},
url = {https://doi.org/10.1371/journal.pone.0359300},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Compact target-aware stability-based feature selection for cross-dataset wearable activity monitoring
AU - Sarra Ayouni
AU - Maha Sliti
AU - Abeer Abdulaziz Alghanem
AU - Fatimah Alhayan
AU - Mohamed Maddeh
PY - 2026
DA - 2026/10/07
JO - PLOS One
DO - 10.1371/journal.pone.0359300
UR - https://doi.org/10.1371/journal.pone.0359300
AB - Wearable activity-recognition models may lose reliability when transferred across participants, sensor placements, sampling rates, acquisition protocols, and activity definitions. This study evaluates a compact unsupervised target-aware feature-selection strategy for cross-dataset wearable activity monitoring under a pre-specified MHEALTH→PAMAP2 transfer protocol. Five-second wearable-signal windows were represented using 270 handcrafted inertial and magnetic descriptors. Source-domain mutual-information relevance was then combined with unlabeled source–target descriptor stability to select a compact set of 48 features, after which lightweight classifiers were trained using MHEALTH labels only. The original activities were harmonized into three coarse states: sedentary, active, and vigorous, while harmonized PAMAP2 target labels were withheld from model development and used only for final evaluation. Under leave-one-subject-out evaluation on MHEALTH, the Random Forest classifier achieved 0.9985 accuracy and 0.9984 macro-F1. In the primary cross-dataset experiment, the pre-specified stability-aware configuration achieved 0.9042 accuracy and 0.8753 macro-F1 on PAMAP2, compared with 0.8888 accuracy and 0.8436 macro-F1 for the full 270-feature representation and 0.8734 accuracy and 0.7940 macro-F1 for source-mutual-information-only selection. With all operating values frozen, reverse PAMAP2→MHEALTH transfer also ranked Stable-48 highest (accuracy 0.5811; macro-F1 0.4400), ahead of Source-MI-48 (0.5720; 0.4320) and Full-270 (0.5445; 0.4257), although Holm-adjusted participant-level contrasts were not significant. The compact configuration also achieved higher complete-window accuracy and macro-F1 than the tested shallow target-feature transformations and feature-level MLP adaptation baselines within the same handcrafted descriptor space. These findings indicate that jointly considering source-domain relevance and unlabeled cross-domain descriptor stability can improve compact cross-dataset wearable activity monitoring. The conclusions remain specific to the evaluated protocols and activity taxonomies and do not establish general superiority over raw-signal deep-learning methods or clinical readiness.
SN - 1932-6203
ER -