مرتضي حدادي

including linear regression, ran‎dom forest, an‎d Long Short-Term Memory (LSTM) networks were implemented to predict actual task durations. The use of multiple models allows for performance comparison an‎d identification of the optimal predictive model capable of capturing complex an‎d nonlinear human productivity patterns. Chapter 4 presents the analysis results. Statistical analysis of the integrated dataset indicates that the average effective activity ratio of personnel is approximately 72%, while 28% of working time is spent on non-productive activities an‎d idle periods. These findings highlight the critical importance of continuous monitoring an‎d practical strategies to reduce idle time an‎d enhance productive activity. Prediction results demonstrate that LSTM networks, with their ability to learn long-term dependencies an‎d capture complex temporal patterns, outperform other models, explaining up to 88% of the variance in actual task durations (R² = 0.88). The ran‎dom forest model also performs well with R² = 0.81, while linear regression serves as a reference with R² = 0.62. Feature importance analysis shows that idle time an‎d effective activity ratio are the most influential factors affecting task durations. Data-driven monitoring of these indicators can significantly reduce project delays an‎d assist in adjusting the critical path. The study demonstrates that combining wearable device data with advanced predictive modeling provides a powerful tool for managing human resource productivity, identifying key factors contributing to delays, an‎d dynamically adjusting project schedules. The proposed framework enables daily performance estimation, critical path identification, real-time material consumption prediction, an‎d schedule adjustments within MSP software. Consequently, project managers can make informed, data-driven decisions to improve workforce efficiency, team coordination, an‎d overall project productivity. The findings not only offer substantial improvements for construction project execution processes but also illustrate the practical applicability of wearable technology an‎d time-series analysis in advanced project management. This data-driven framework is adaptable to real-world projects with sequential an‎d parallel activities under variable site conditions, serving as a foundation for developing intelligent predictive an‎d schedule management systems in construction projects.

بهره ‌وري نيروي كار، حسگرهاي پوشيدني، داده‌هاي حركتي، يادگيري ماشين، اصلاح تطبيقي زمان‌بندي، مسير بحراني، مديريت پروژه‌هاي ساختماني

Human productivity, wearable sensors, motion data, machine learning, adaptive schedule adjustment, critical path, construction project management

Morteza hadadi

Dr. Aliakbar Shirzadi Javid