War Prediction Model

🎯 War Prediction Model

📊 Research Project: ML-based conflict probability prediction
Author: Oleh Ivchenko, PhD Candidate
Version: 0.2.0 (Updated)
References: Hegre et al. 2019, Mueller & Rauh 2018

Overview

This project uses machine learning to predict the probability of armed conflict for any country, combining multiple data sources and academic research on conflict prediction.

📄 Academic Research Paper Published!
Predicting Armed Conflict Probability: A Multi-Factor Machine Learning Approach
Ivchenko, O. (2026). Spec-Driven AI Development Series. Odessa National Polytechnic University.
DOI: 10.5281/zenodo.18735965
📖 Read Full Paper

📓 Technical Documentation

📓

Jupyter Notebook: War Prediction Model

Complete technical implementation with code examples, model architecture, training process, and usage demonstrations. Includes data preprocessing, feature engineering, ensemble modeling, and SHAP explainability.

📚 What’s Inside:
  • Data loading from ACLED, UCDP, World Bank, SIPRI, V-Dem
  • Feature engineering (economic, political, military indicators)
  • Ensemble model (XGBoost + Random Forest + LSTM + Logistic Regression)
  • Model training with cross-validation
  • Performance evaluation and metrics
  • SHAP explainability analysis
  • Example predictions for multiple countries
  • Complete code with documentation

🗺️ World Conflict Risk Map

Interactive visualization showing conflict probability by country. Click any country or use the dropdown to see detailed risk analysis.

Loading map…

Model Architecture

The prediction model uses an ensemble of machine learning algorithms trained on historical conflict data:

flowchart TB
    subgraph Input["📥 Input Features"]
        E[Economic Indicators]
        P[Political Stability]
        M[Military Activity]
        G[Geographic Factors]
        H[Historical Conflicts]
    end
    
    subgraph Model["🤖 Ensemble Model"]
        XG[XGBoost]
        RF[Random Forest]
        LSTM[LSTM Temporal]
        LR[Logistic Regression]
    end
    
    subgraph Output["📊 Output"]
        PROB[Probability 0-1]
        CONF[Confidence Score]
        FACTORS[Factor Weights]
    end
    
    E --> XG & RF & LSTM & LR
    P --> XG & RF & LSTM & LR
    M --> XG & RF & LSTM & LR
    G --> XG & RF & LSTM & LR
    H --> XG & RF & LSTM & LR
    
    XG & RF & LSTM & LR --> PROB
    XG & RF & LSTM & LR --> CONF
    XG & RF & LSTM & LR --> FACTORS

Data Sources

SourceData TypeCoverage
ACLEDArmed conflict eventsGlobal, real-time
GDELTGlobal events databaseGlobal, historical
World BankEconomic indicators195 countries
SIPRIMilitary expenditureGlobal, annual
UCDPConflict data1946-present
V-DemDemocracy indices202 countries

Academic References

  1. Hegre, H., et al. (2019). “Predicting Armed Conflict, 2010-2050.” Journal of Peace Research, 56(4). DOI: 10.1177/0022343319834048
  2. Mueller, H., & Rauh, C. (2018). “Reading Between the Lines: Prediction of Political Violence.” American Political Science Review, 112(2). DOI: 10.1017/S0003055417000533
  3. Ward, M. D., et al. (2013). “Learning from the Past and Predicting the Future.” International Interactions, 39(3). DOI: 10.1080/03050629.2013.782306
  4. Muchlinski, D., et al. (2016). “Comparing Random Forest with Logistic Regression.” Political Analysis, 24(2). DOI: 10.1093/pan/mpv024

This is a research project by Stabilarity Research Hub. Predictions are based on publicly available data and academic models. Not financial or political advice.