A Predictive Logistic Regression Model of World Conflict Using Open Source Data
Abstract
Nations transitioning into conflict is an issue of national interest. This study considers various data for inclusion in a statistical model that predicts the future state of the world where nations will either be in a state of "violent conflict" or "not in violent conflict" based on available historical data. Logistic regression is used to construct and test various models to produce a parsimonious world model with 15 variables. Further analysis shows that nations differ significantly by geographical area. Therefore six sub-models are constructed for differing geographical areas of the world. The dominant variables for each sub-model vary, suggesting a complex world that cannot be modeled as a whole. Insights and conclusions are gathered from the models, a best model is proposed, and predictions are made for the state of the world in 2015. Accuracy of predictions via validation surpasses 80%. Eighty-five nations are predicted to be in a state of violent conflict in 2015, seventeen of them are new to conflict since the last published list in 2013. A prediction tool is created to allow war-game subject matter experts and students to identify future predicted violent conflict and the responsible variables.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 26, 2015
- Accession Number
- ADA615064
Entities
People
- Benjamin C. Boekestein
Organizations
- Air Force Institute of Technology