Forecasting Military Conflicts Using Predictive Analytic Models

Lead Author Major

Political Science

Lead Author Status

Senior

Format

Oral Presentation

Faculty Mentor Name

Dari Sylvester Tran

Faculty Mentor Department

Political Science

Abstract/Artist Statement

Meticulous analysis of the foreign policy, socioeconomic conditions, and leader’s psychology, among numerous other factors, is often performed in order to determine the causal mechanisms of war. All in order to understand war as something more than a random and sporadic event. I question whether war can be predicted. War is defined for the purpose of this research as a military conflict between two parties which results in at least twenty-five battle-related deaths (UCDP). This paper explores the literature around predictive modeling and evaluates existing predictive analytic models of war. I analyze models that use science, statistics, machine-learning techniques, and even literature to predict military conflicts. There have been various forecasts established, both for the next few years and decades into the future. As a criterion for analyzing the effectiveness of the models, I use the Russian-Ukrainian war as a case study. Where applicable, if the model accurately predicted the current war, then it is a useful predictive model. I then attempt to build my own model using classification algorithms and machine learning to examine interstate conflicts in the coming years using the UCDP/PRIO dataset of armed conflicts from 1946-2020 and relevant indicators from the World Bank. Predicting wars and understanding them go together, our knowledge of one enhances that of the other. If we can better predict where military conflict might arise next, we can better act proactively rather than respond to the aftermath of war.

Location

Yosemite Learning Lab, William Knox Holt Memorial Library and Learning Center

Start Date

30-4-2022 1:40 PM

End Date

30-4-2022 1:59 PM

This document is currently not available here.

Share

COinS
 
Apr 30th, 1:40 PM Apr 30th, 1:59 PM

Forecasting Military Conflicts Using Predictive Analytic Models

Yosemite Learning Lab, William Knox Holt Memorial Library and Learning Center

Meticulous analysis of the foreign policy, socioeconomic conditions, and leader’s psychology, among numerous other factors, is often performed in order to determine the causal mechanisms of war. All in order to understand war as something more than a random and sporadic event. I question whether war can be predicted. War is defined for the purpose of this research as a military conflict between two parties which results in at least twenty-five battle-related deaths (UCDP). This paper explores the literature around predictive modeling and evaluates existing predictive analytic models of war. I analyze models that use science, statistics, machine-learning techniques, and even literature to predict military conflicts. There have been various forecasts established, both for the next few years and decades into the future. As a criterion for analyzing the effectiveness of the models, I use the Russian-Ukrainian war as a case study. Where applicable, if the model accurately predicted the current war, then it is a useful predictive model. I then attempt to build my own model using classification algorithms and machine learning to examine interstate conflicts in the coming years using the UCDP/PRIO dataset of armed conflicts from 1946-2020 and relevant indicators from the World Bank. Predicting wars and understanding them go together, our knowledge of one enhances that of the other. If we can better predict where military conflict might arise next, we can better act proactively rather than respond to the aftermath of war.