MAFEM: A Fuzzy Logic-Based Multi-Attribute Evaluation Framework for Personalized Physical Activity Monitoring Using Wearable Sensors

Abstract

The integration of physical activity into students' daily routines is vital for improving both health and academic outcomes. To ensure the effectiveness of such activities, it is crucial to implement systems that can accurately monitor and analyze exercise data. This study proposes the Multi-Attribute Fuzzy Evaluation Model (MAFEM), which utilizes fuzzy logic and fuzzy sets to interpret complex sensor data and assess student health. The model incorporates preprocessing, fuzzification, defuzzification, and rule evaluation, optimized through adaptive thresholds for enhanced personalization. MAFEM was evaluated using the MM-Fit dataset, which includes synchronized multimodal data (e.g., accelerometer, gyroscope, and heart rate) collected from 30 university students performing standardized physical activities (walking, jogging, cycling, stair climbing, and resting). The system was benchmarked against three state-of-the-art methods: SHER, HAD, and HNN frameworks. Experimental validation involved 50 exercise sessions in both indoor and outdoor environments, with performance metrics computed based on standard evaluation protocols. MAFEM demonstrated high reliability, achieving 97.11% precision, 95.84% recall, and an RMSE of 0.23. Furthermore, it maintained low computational complexity (O(r.m.n)) and minimal energy consumption (approximately 65mAh during Wi-Fi-based operation), outperforming baseline models in both accuracy and resource efficiency. These findings highlight the robustness and practicality of fuzzy logic-driven multi-attribute frameworks for personalized, real-time physical activity monitoring in wearable health systems.