/amnesty_logit_svm_knn

Primary LanguageR

Comparing Logistic Regression, SVM, kNN for predicting follow-up in all Amnesty International Urgent Actions (not split by category, yet)

save_dtm.R

  • Run once to generate ./data/amnesty_dtm.RData
    • DTM of Urgent Actions
    • 99% sparse: ~5165 Urgent Actions, with 3506 inputs from DTM
    • 95% sparse (./data/amnesty_s95_dtm.RData): ~5165 Urgent Actions, with 653 inputs from DTM

amnesty_model_compare.R

  • Compares:
    • Logistic Regression w/ LASSO regularization via 10-fold X-valiation
    • kNN (k = 5-50)
    • SVM (linear,poly,radial) w/ cost = 1-10