Title

Classification Trees with Synthetic Features for Multiclass Classification Problems

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Mathematics

Date of Award

Spring 2022

Abstract

In data mining, classification is considered one of the problems under supervised machine learning. One popular method for solving classification problems is the classification tree. Most of the algorithms for classification tree are used for binary classifiers. We throw more light on binary classification in this work and extend the knowledge into a multiclass classification problems. The two main approaches for dealing with multiclass classifications are reducing the multiclass classification problem to a collection of binary classification problems and directly boosting the multiclass classification problems. We give notes on the concepts of multiclass classification problems, throw more light on the different methods that are under the two main approaches, use the first method to conduct experiments with reliable datasets, then use the concept of cross-validation to fit, test and compare the accuracy of different models. We make use of original scripts in R and use concepts such as variable importance, and voting when fitting model and making predictions.

Advisor

Thomas Boucher

Subject Categories

Applied Mathematics | Mathematics | Physical Sciences and Mathematics

Share

COinS