Academic Commons

Theses Doctoral

Modeling the Likelihood of Construction Incidents Using Public Data

Gerstenberger, Armand

There has been an upward trend of construction injuries and fatalities in the recent decade. Regulatory agencies, such as the NYC Department of Buildings, exist to create and modify construction safety laws, review construction projects, and enforce these laws through site inspections, and often make the data they collect available to the public. However, there is a lack of predictive modeling and a lack of research regarding how to make a proactive prediction of potential injuries and fatalities on construction sites. This study uses public data to predict future construction incidents using leading indicators from information gathered from the NYC permits-issued and complaints-received databases. Results indicate that it is possible to predict future construction incidents over multiple forecast windows using a logistic regression and zero-inflated Poisson model. While previous site incidents alone are significant in predicting future incidents, adding permit and complaint related information increased the true positive rate and decreased the false-negative rate.

Geographic Areas

Files

  • thumnail for Gerstenberger_tc.columbia_0055E_11144.pdf Gerstenberger_tc.columbia_0055E_11144.pdf application/pdf 2.31 MB Download File

More About This Work

Academic Units
Human Development
Thesis Advisors
Corter, James E.
Keller, Bryan
Degree
Ed.D., Teachers College, Columbia University
Published Here
February 23, 2021