Envelope-based support vector machines classification

Alzahrani, Alya 2022. Envelope-based support vector machines classification. PhD Thesis, Cardiff University. Item availability restricted.

Preview	PDF - Accepted Post-Print Version Download (2MB) | Preview
	PDF (Cardiff University Electronic Publication Form) - Supplemental Material Restricted to Repository staff only Download (123kB)

Abstract

Envelope methodology is a promising dimension reduction approach. It was introduced in the regression framework. In this work, we extended envelope application and focused on the reduce-and-classify approach in supervised learning. The first contribution is that we extended this method to classification and developed a new projection-based approach based on a Support Vector Machine (SVM) classifier. Our proposed classifier ESVM (Envelope-based Support Vector Machines) is obtained by combining the envelope method and SVM to achieve a better and more efficient classification. Using the idea of the envelope to extract a lower-dimensional subspace projected the data on has advanced the classification performance. The empirical results show a low misclassification rate based on ESVM Furthermore, we extended the ESVM classifier to sparse data. In that, the reducing subspace reduces the dimension and selects significant variables simultaneously. We employ an adaptive group lasso penalty to impose the sparsity in the reducing subspace. The classifier is evaluated based on simulation and real data.

Item Type:	Thesis (PhD)
Date Type:	Completion
Status:	Unpublished
Schools:	Advanced Research Computing @ Cardiff (ARCCA) Mathematics
Date of First Compliant Deposit:	21 February 2023
Last Modified:	01 Aug 2024 10:10
URI:	https://orca.cardiff.ac.uk/id/eprint/157203

Actions (repository staff only)

Edit Item