AI & Statistics Seminar: Test Strategies for Cost-Sensitive Decision Trees

Speaker:	Prof. Charles Ling
		University of Western Ontario, Canada

Topic:		"AI & Statistics Seminar: Test Strategies for
		 Cost-Sensitive Decision Trees"

Date:		Tuesday, 1 June 2005

Time:		10:30 am - 11:30am

Venue:		Room 3464 (Joint Conference, via lift nos. 25/26)
		HKUST

Abstract:

In medical diagnosis doctors must often determine what tests (e.g., blood
tests) should be ordered for a patient to minimize the total cost of tests
and misclassifications (misdiagnosis). In this talk, we model this
learning and test process that incorporates both test costs and
misclassification costs. In particular, we first propose a lazy decision
tree learning algorithm that minimizes the total cost of tests and
misclassifications. Then we design several novel test strategies which
attempt to minimize the total cost for new test examples (new patients).
We empirically evaluate our strategies, and show that they are very
effective. Our results can be readily applied to real-world diagnosis
tasks, such as medical diagnosis. A case study on heart disease is given.
If time allows, we will also talk about the role of missing values in
cost-sensitive learning.

This is a joint work with Drs. Qiang Yang, Shichao Zhang, and Shengli Sheng.
Parts of the work were published in ICML'04 and IEEE TKDE (2005).



***********************
Biography:

Professor Charles Ling earned his Msc and PhD from the Department of
Computer Science at Univ of Pennsylvania in 1987 and 1989 respectively.
Since then he has been a faculty member in Computer Science at University of
Western Ontario. His main research areas include machine learning, data
mining, and cognitive modeling. He has published over 100 research papers in
journals (such as Machine Learning, JAIR, JMLR, JKDD, IEEE TKDE, and
Bioinformatics) and international conferences (such as IJCAI, KDD, ICDM, and
ICML). He is also the Director of Data Mining Lab, leading data mining
development in CRM, Bioinformatics, and the Internet. He has managed several
major data-mining projects for banks and insurance companies in Canada. See
http://www.csd.uwo.ca/faculty/cling for more info.