Related Experiment Video
Updated: Jul 11, 2025
06:37
Author Spotlight: Addressing Technical and Subjective Challenges in Measuring Classroom Attention
Published on: December 15, 2023
3.8K
Predicting Office Workers' Productivity: A Machine Learning Approach Integrating Physiological, Behavioral, and
Mohamad Awada1, Burcin Becerik-Gerber1, Gale Lucas2
1Department of Civil and Environmental Engineering, University of Southern California, Los Angeles, CA 90089, USA.
Sensors (Basel, Switzerland)
|November 14, 2023
Summary
This study uses machine learning to predict office worker productivity by analyzing physiological, behavioral, and psychological data. Incorporating psychological states significantly improved prediction accuracy, highlighting mood and eustress as key factors.
Area of Science:
- Human-Computer Interaction
- Machine Learning Applications
- Occupational Health Psychology
Background:
- Office worker productivity is crucial for organizational success.
- Accurate assessment of productivity is challenging due to its subjective nature.
- Understanding the interplay of physiological, behavioral, and psychological factors is key to improving productivity.
Purpose of the Study:
- To develop and evaluate a machine learning framework for predicting perceived office worker productivity.
- To compare the predictive performance of models incorporating only physiological and behavioral data versus those including psychological states.
- To identify key predictors of productivity and assess the efficacy of different data collection methods.
Main Methods:
- Utilized various machine learning models, including XGBoost, to predict psychological states and productivity.
- Compared a baseline model (physiological, behavioral data) with an extended model (including psychological states).
- Analyzed feature importance and compared data from wearable devices (Empatica E4, H10 Polar) against workstation add-ons (Kinect, computer usage monitoring).
Main Results:
- The extended model significantly outperformed the baseline model, achieving R²=0.60 and MAE=10.52 versus R²=0.48 and MAE=16.62.
- Mood and eustress were identified as significant predictors of productivity, alongside physiological and behavioral features.
- Wearable devices demonstrated superior performance in predicting productivity compared to workstation add-ons.
Conclusions:
- Machine learning frameworks integrating psychological states enhance the prediction of office worker productivity.
- Wearable devices offer a promising approach for independent productivity assessment.
- Smart workstations can leverage these models to create adaptive environments promoting productivity and well-being.
