方法对比

	Machine Learning Conflict Prediction	Conflict Forecasting
领域	International Relations	International Relations
方法族	Machine learning	Machine learning
起源年份	2016	2019
提出者	Predictive conflict research (e.g., Muchlinski, Siroky, He & Kocher)	Conflict-forecasting community (e.g., Håvard Hegre and the ViEWS team)
类型	Supervised machine-learning prediction of conflict	Operational predictive system for armed conflict
开创性文献	Muchlinski, D., Siroky, D., He, J., & Kocher, M. (2016). Comparing random forest with logistic regression for predicting class-imbalanced civil war onset data. Political Analysis, 24(1), 87–103.	Hegre, H., Allansson, M., Basedau, M., Colaresi, M., Croicu, M., Fjelde, H., et al. (2019). ViEWS: A political violence early-warning system. Journal of Peace Research, 56(2), 155–174.
别名	ML Conflict Prediction, Random Forest Civil War Prediction, Algorithmic Conflict Prediction, Supervised Learning for Conflict Onset	Political Violence Early Warning, Armed Conflict Forecasting, Conflict Early-Warning Systems, ViEWS-Style Forecasting
相关	3	3
摘要	Machine learning conflict prediction uses flexible supervised algorithms — random forests, gradient boosting, neural networks, regularized regression — to forecast the onset of armed conflict from large sets of features, prioritizing out-of-sample predictive accuracy over coefficient interpretation. Muchlinski, Siroky, He, and Kocher (2016) showed that random forests substantially outperform logistic regression at predicting class-imbalanced civil-war onset, catalyzing a shift in conflict research toward algorithmic prediction, rigorous out-of-sample validation, and the recognition that explanation and prediction are distinct goals.	Conflict forecasting is the enterprise of producing calibrated, regularly updated probabilistic predictions of where and when armed conflict will occur, to support early warning and prevention. Exemplified by operational systems such as ViEWS (Hegre et al. 2019), it combines historical conflict data and predictors at fine spatial and temporal resolution, fits and ensembles multiple models, and forecasts violence months ahead — then rigorously evaluates those forecasts against what actually happens. It differs from explanatory conflict analysis by being transparent, prospective, and judged on out-of-sample accuracy rather than on coefficients.