Project 3: Adaboost Algorithm

Boosting is an additive model, but is different from generalized additive model, in which each weak learner only involves one variable and p number of functions are used and added up. Boosting is also different from random forests, another additive model. In random forests, each tree is generated independently, so they can’t borrow information from each other. Adaboost is a special case of this framework with exponential loss for classification.

July 26, 2020

Project 1: KNN classification

K-Nearest Neighbour (KNN) is a nonparametric method that predicts the target point x with averages of nearby observations in the training data. The algorithm is simple and effective. It makes no assumptions about the underlying data distribution. However, it does not produce a model, which limits the ability to find new insights in relationships among features. In this project I wrote my own KNN algorithm which I compared to the model already available in R.