The Geography of Genocide: Using Machine Learning to Locate Undocumented Mass Graves of the Holocaust
Open Access
- Author:
- Schierman, Kelly
- Graduate Program:
- Spatial Data Science
- Degree:
- Master of Science
- Document Type:
- Master Thesis
- Date of Defense:
- September 30, 2024
- Committee Members:
- Brandi Gaertner, Thesis Advisor/Co-Advisor
Leanne Sulewski, Committee Member
Anthony Robinson, Program Head/Chair - Keywords:
- MaxEnt
machine learning
holocaust
mass graves
lithuania
holocaust by bullets - Abstract:
- Following their invasion of the Soviet Union in 1941, the Germans killed at least two million Jews in Eastern Europe by shooting them at close range and burying them in pits. Thousands of these mass graves are scattered across at least ten countries, and an unknown number of graves remain today forgotten and undocumented. The most commonly used methods of locating undocumented mass graves involve employing a combination of archival research, surface-based archaeological investigations, and, perhaps most critically, the testimony of eyewitnesses or survivors. However, the advanced age of most witnesses has created an imperative for developing new methods to support the discovery of undocumented mass graves. The maximum entropy (MaxEnt) machine learning algorithm uses a series of environmental variables to describe known presence points – in this case, the locations of mass graves – and that information is used to generate a geospatial model for predicting possible presence locations for undocumented mass graves. The objective of this study was to use MaxEnt to develop a viable model for predicting potential presence locations for undocumented mass graves as an assist to traditional methods. The most successful predictive model, using a combination of 13 geospatial variables, provided an area under the curve (AUC) value of 0.9417 and an omission rate of 0.0690, correctly identifying the locations for 207 out of a total of 229 known mass graves.