The College of Design congratulates Sungil Hong for the successful defense of their dissertation, "Analyzing Physical Workplace and Service Management Using Natural Language Processing and Machine Learning Approaches" on Friday, April 1st, 2022.

Abstract:

Analyzing Physical Workplace and Service Management Using Natural Language Processing and Machine Learning Approaches

The demand for workplace flexibility has emerged according to ever-changing environments, such as sharing and gig economy, alternative work arrangement, and COVID-19. This study proposes a redefined facility management model corresponding to the changing circumstances, which provides not only space but also activity support and leisure services. Coworking space (CWS) is one of the embodiments of the model. This research aims to develop CWS management strategies for 1) user preferences in physical workplace environments and services during COVID-19 and 2) data management methods utilizing natural language processing (NLP) and machine learning techniques. Two main studies in this research address three research objectives: 1) identifying preferences for facilities and services factors in CWSs during COVID-19; 2) detecting changing preferences for factors about facilities and services during COVID-19; 3) proposing the applications of machine learning and NLP techniques and demonstrating the applicability of computational data collection and analysis methods in the physical workplace management research. First, Study I proposes a thematic categorization scheme of CWS spatial and service factors and elements. Based on the categories, a mixed-method approach was utilized for the comprehensive data analysis, including content analysis, classification, and clustering. The results show that CWS users have become sensitive to disruptive behaviors and hygienic responses to infectious diseases after the pandemic. The findings also present a need for a sense of community and various technology needs for virtual interactions. Second, Study II performed the data integration of a large computerized maintenance management system dataset of a public college campus into a single CWS building maintenance dataset to build robust machine learning-based text classification models for a small dataset. The results show the qualitative and quantitative increase in prediction performance of text classifications. Study II implies that data integration will accelerate smart facility management, including small or single buildings, by sharing public datasets. In conclusion, this research sheds light on online big data collection and analysis in physical workplace management research. It also presents how the facility management industry can apply such state-of-the-art technology in utilizing historical data to make data-driven decisions.