Skip to content

Conflict Forecasting via Conformal Prediction for Markov Processes

Home / Papers / Conflict Forecasting via Conformal Prediction for Markov Processes

Conflict Forecasting via Conformal Prediction for Markov Processes

Authors: Aditya Basarkar, Emmett B. Kendall, David Randahl, Jonathan P. Williams, Gudmund H. Hermansen Date: 2026-04-28 Paper ID: openalex:2604.25139

Summary

This paper addresses the challenge of forecasting conflict dynamics by employing conformal prediction to generate valid uncertainty sets for future state-sequences in Markov processes. Unlike traditional point-prediction methods, which may carry extreme risks in geopolitical contexts, this approach provides statistically valid coverage even under potential model misspecification. The authors evaluate the methodology against likelihood-based strategies and offer a critical discussion on the theoretical hurdles of applying conformal prediction to non-exchangeable temporal data.

Key Contributions

  • Introduces a conformal prediction framework for generating valid uncertainty sets for future conflict state-sequences in Markovian processes.
  • Provides a robust alternative to likelihood-based prediction strategies, addressing risks of model misspecification in high-stakes conflict forecasting.
  • Critically analyzes the applicability of standard conformal prediction methods to non-exchangeable, temporally-dependent Markovian data.

Metadata & Links

url
https://arxiv.org/abs/2604.25139
paper_id
2604.25139
paper_source
openalex
domain
time-series
tags
uncertainty-quantificationconformal-predictionmarkov-modelsconflict-forecasting
architectures
datasets
concept_slugs
dataset_slugs
skill
TimeSeriesSkill
created_at
2026-05-01T06:04:39Z