Term of Award

Spring 2025

Degree Name

Master of Science, Civil Engineering

Document Type and Release Option

Thesis (open access)

Copyright Statement / License for Reuse

Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.

Department

Department of Civil Engineering and Construction

Committee Chair

Myung Jeong

Committee Member 1

Soonkie Nam

Committee Member 2

Matthew Ricks

Abstract

Construction safety incidents remain a significant concern, particularly in the Southeastern U.S. due to the high-risk nature of the industry. Analyzing patterns in these incidents can help improve safety practices and reduce accidents. Machine learning (ML) techniques were employed in this study to identify temporal patterns in construction safety incidents, aiming to enhance proactive safety management. The machine learning methods used in this research included logistic regression, decision trees, random forest, support vector machine (SVM), and K nearest neighbors (KNN).

The objective of the study was to analyze temporal trends in safety incidents and identify the most effective machine learningtechnique for predicting and classifying these incidents. The data used in this study was sourced from the Occupational Safety and Health Administration (OSHA) database, focusing on safety incidents in the Southeastern U.S. construction industry from 2013 to 2023. A total of 1,963 cases were analyzed, and categorized as fatal, hospitalized, and non-hospitalized. Each case included detailed project and injury information. The analysis was conducted using Python and Jupyter Notebook, with separate notebooks created for each machine learning technique to streamline the coding and evaluation process. Three main questions were explored: 1) Which machine learning technique offers the most accurate prediction in classifying between fatal and non-fatal incidents? 2) How can machine learning models be used to assess the likelihood of fatal and non-fatal incidents? 3) What are the key temporal patterns (daily, seasonal, and yearly) observed in construction safety incidents in the Southeastern U.S.? The findings showed that random forest and decision trees were the most effective in predicting safety incidents, with random forest achieving the highest accuracy and reliability for both fatal and non-fatal classifications. This study highlights the potential of machine learning in improving construction safety by offering more accurate predictions and insights into high-risk incidents, aiding in better decision-making and risk management strategies.

OCLC Number

1520497627

Research Data and Supplementary Material

Yes

Download

Share

COinS