Author

Sarah Neumann

Date of Award

3-2018

Document Type

Thesis

Degree Name

Master of Science in Operations Research

Department

Department of Operational Sciences

First Advisor

Darryl K. Ahner, PhD

Abstract

Conflict forecasts are crucial to Combatant Commanders’ understanding of the dynamic environment encompassing countries within their area of responsibility. The current structure of the Combatant Commands (COCOMs) is rooted in geography by grouping nations in geographic proximity to the same regional command. However, leaders today question the effectiveness of the current structure. A novel modified k-means clustering algorithm is developed and implemented that groups countries based on data similarities and geographic proximity resulting in new COCOM groupings that improve conflict forecasts. The data spans various political, military, economic, and social characteristics of countries, and is used to develop conditional logistic regression models that predict the likelihood of a country to transition into or out of conflict. Predictions from models using the original COCOM regions are compared to predictions from models using the new, data and geography based, regions. Reorganizing the COCOMs based on data similarity and geography improves conflict forecasting models and prediction capabilities significantly. The new models achieved training data classification accuracies exceeding 89% for each new COCOM and improved predicting the validation data with classification accuracies increasing up to 2% as compared to the current COCOM models and up to 2.5% in comparison to previous conflict studies. This methodology of grouping countries based on data similarities and geographic proximity leads to newly defined COCOMs which improve overall forecasting ability of conflict transitions, the best found in the literature to date. These results can help Combatant Commanders better understand the dynamic conflict environment of their area of responsibility and lead to the development of more effective operational and strategic plans.

AFIT Designator

AFIT-ENS-MS-18-M-149

DTIC Accession Number

AD1056378

Share

COinS