Statistics for Data Science

Ready for boosting

There is now a function able to train our ensemble classifier using AdaBoost, but we also need a function to make the actual predictions. This function will take in the output list produced by our training function, AdaBoostNN(), along with a test dataset.

This function is AdaBoostNN.predict() and it is shown as follows:

AdaBoostNN.predict <- function(ada_model, test_data) { 
  models <- ada_model$models 
  alphas <- ada_model$alphas 
  prediction_matrix <- sapply(models, function (x)  
             as.numeric(predict(x, test_data, type = "class"))) 
  weighted_predictions <- t(apply(prediction_matrix, 1,  
             function(x) mapply(function(y, z) y * z, x, alphas))) 
  final_predictions <- apply(weighted_predictions, 1,  function(x) sign(sum(x))) return(final_predictions) ...