Random Forests : An Application To Tumour Classification

Abstract

In this thesis, machine learning approaches, namely decision trees and random forests, are discussed. A mathematical foundation of decision trees is given. It is followed by discussion of the advantages and disadvantages of them. Further, the application of decision trees as a part of random forests is presented. A real life study of brain tumours is discussed regarding usage of random forests. The data consists of six different types of brain tumours, and the data is acquired by Raman spectroscopy. After the data has been curated, a random forest model is utilised to classify the class of the tumour. At the current point, the results seem optimistic, but require further experimentation.