Prevention Is Better Than Cure: Machine Learning Approach to Conflict Prediction in Sub-Saharan Africa

This article offers policymakers and researchers pragmatic sustainable approaches to identify mitigate conflict threats by looking beyond p-values plausible instruments. We argue that predicting successfully depends on the choice of algorithms, which, if chosen accurately, can reduce economic social instabilities caused post-conflict reconstruction. After collating data with variables linked conflict, we used a grid level dataset 5928 observations spanning 48 countries across sub-Saharan Africa predict civil conflict. The goals study were assess performance supervised classification machine learning (ML) algorithms in comparison logistic model, implication selecting specific metric policy initiatives, evaluate value interpretability selected model. comparing class imbalance resampling methods, synthetic minority over-sampling technique (SMOTE) was employed improve out-of-sample prediction for trained results indicate our is recall, gradient tree boosting best algorithm; however, precision or F1 score metric, then multilayer perceptron algorithm produces