Quality of Classification Approaches for the Quantitative Analysis of International Conflict
We provide an evaluative comparison of some modern classification algorithms, such as CART, AdaBoost, bagging and random forests, to predict the incidences of military conflicts and other political relevant events. Our evaluative comparison is based on two main aspects: the importance of variables within the classifier as well as the prediction accuracy. While modern classification procedures are able to improve the prediction accuracy as compared to the traditionally used logistic regression, the logistic regression still holds a large advantage in terms of interpretability of the variables’ relevancy.
KeywordsLogistic regression Classification trees Boosting Rare events
- 8.Palmer, G., D’Orazio, V., Kenwick, M., Lane, M.: The mid4 data set: Procedures, coding rules, and description. Confl. Manag. Peace Sci. 32(2), 222–242 (2015)Google Scholar
- 9.R Core Team: R—A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria (2014)Google Scholar