Understanding and Predicting Sit-Stand Desk Usage Patterns and Willingness among Knowledge Workers: A Data-Driven Approach

Abstract

This research was conducted in two distinct phases to investigate and forecast sit-stand desk usage among knowledge workers. In Phase 1, we performed an observation study to collect desk height and contextual data from the workers and analyzed the primary factors influencing a worker's willingness to switch postures. Our analysis revealed key contextual features that are critical determinants of ergonomic behavior, providing a deeper understanding of the interplay between environmental and behavioral factors in sit-stand desk usage. In Phase 2, we developed a time-series predictive system that integrates an XGBoost model with a cluster-based customization for forecasting workers' intention to stand as well as their actual work postures. This framework tailors predictions to the unique characteristics of different user groups, resulting in enhanced forecasting accuracy and smoother, less noisy predictive outputs by focusing on recurring behavioral patterns. With the customization, we were able to forecast the intention of the user with 0.05 mean squared error and posture of the user with 99% of accuracy. Future work will explore adaptive nudging strategies to optimize the timing and frequency of alerts, further promoting healthy and productive work habits.