Project Title: Data Mining Using Support Vector Machines
Adviser: Dr. Hickman
Description: Support vector machines (SVM) have become
increasingly popular tools for data mining tasks such as data classification. A
SVM uses optimization algorithms to determine the appropriate classification
based on characteristics of the input. Specifically, the input data is mapped
to another vector space and the SVM is designed to identify linear relations in
the new space. Linear programming is used iteratively to separate the data into
various classes.
Project
requirements:
Study and understand the paper: Bennett &
Campbell, 2000, Support Vector Machines: Hype or Hallelujah?, SIGKDD
Explorations 2:2, pp 1-13.
Study
additional resources including:
Cristianini & Shawe-Taylor, An
Introduction to Support Vector Machines and Other Kernel-based Learning
Methods, 2000, Cambridge University Press.
Schkopf & Smola, Learning with
Kernels: Support Vector Machines, Regularization, Optimization, and Beyond,
2001, MIT Press.
Implement a SVM in Maple (or similar language)
and test your SVM code on standard data classification problems from existing
test suites (computational emphasis). Or adapt and apply existing public domain
SVM software to solve classification problems in a new domain such as medicine,
criminal justice, or remote sensing (application emphasis).
Prepare a report to presented
at the UNO Math Department MAM Symposium in April 2005.
Students interested in this project should have
completed or be currently enrolled in MATH 4300 and be comfortable working in a
computational environment.