Office Space and Gentrification in King County: A Machine Learning Based Approach

Abstract

In recent years, King County has seen a surge in housing costs and unaffordability, leading to a housing crisis. Local newspapers and community groups have pointed to large companies and the highly paid employees they attract as responsible for unaffordability and gentrification in the area. Though intuitively this may appear to be the case, this thesis uses newly available Machine Learning (ML) technology to quantitatively investigate the importance of offices as they correlate with gentrification. Making use of data made available through the American Community Survey and King County GIS, patterns of gentrification examined from 2010 to 2019. Following this, demographic and aggregated office-related variables are established at the block group level, and new GPU-boosted methods of performing ML and SHAP analysis are used to investigate the level to which office-related variables, such as taxable land value, office age, etc., are correlated with a prediction of gentrification within a block group.