Data classification is an archetypal problem in data science. A common approach is the use of decision trees, which takes a variety of forms. This includes single tree classifiers such as oblique trees like HH-CART (= Householder Classification And Regression Trees), as well as weak classifiers like stumps (trees with two leaves) which can be used in ensemble methods. In practice methods like HH-CART grow a large tree branch by branch, and then prune it to obtain the final classifier. This process of generating a branch is closely related to that of generating a stump. This project will look at how stumps and depth two trees can be generated via global optimization techniques, and how they can be employed in single tree and ensemble methods.
Supervisors
Primary Supervisor: Chris Price
Other Supervisors: Blair Robertson, Marco Reale
Key qualifications and skills
Bachelors in mathematics, statistics, data science, or similar.
Does the project come with funding
No - Student must be self-funded
Final date for receiving applications
Ongoing
How to apply
Apply by email to primary supervisor with CV etc.
Keywords
Data classification; decision trees; global optimization
