ICML-2000 Workshop on COST-SENSITIVE LEARNING
SCHEDULE (July 2, 2000)
1. Introduction to Cost-Sensitive Learning
- 9:00-9:50 Foster Provost: Tutorial Overview:
The Cost-Sensitive Learning Problem(s) and a Review of the Main Existing Approaches to Solving Them.
- 9:50-10:20 Questions and Discussion
10:20-10:40 BREAK
2. Algorithms for Cost-Sensitive Learning
- 10:40-11:10 Pedro Domingos: Research Directions in MetaCost
- 11:10-11:30 Questions and Discussion
- 11:30-11:45 Charles Elkan: Cost-Sensitive Learning and
Decision-Making When Costs Are Unknown
- 11:45-12:00 Questions and Discussion
- 12:00-12:15 Joao Gama: A Cost-Sensitive Iterative Bayes
- 12:15-12:30 Questions and Discussion
12:30-14:00 LUNCH
3. Real-World Loss Functions and Types of Costs
- 14:00-14:20 Peter Turney: Types of Cost in Inductive Concept Learning
- 14:20-14:30 Questions and Discussion
- 14:30-14:45 Kai Ming Ting: Cost-Sensitive Classification using
Decision Trees, Boosting and MetaCost
- 14:45-15:00 Questions and Discussion
- 15:00-15:15 Feng Yu Lin & Sally McClean: The Prediction of
Financial Distress using a Cost Sensitive approach
and Prior Probabilities
- 15:15-15:30 Questions and Discussion
15:30 - 15:45 BREAK
4. Representing Costs and ROC Analysis
- 15:45-16:00 Terran Lane: Extensions of ROC Analysis to Multi-Class Domains
- 16:00-16:15 Questions and Discussion
- 16:15-16:30 Chris Drummond & Robert Holte: Explicitly Representing
Expected Cost: An Alternative to ROC Representation
- 16:30-16:45 Questions and Discussion
5. Methodology and Future Research
- 16:45-17:00 Dragos Margineantu: On Class-Probability Estimates and Cost-Sensitive Evaluation of Classifiers
- 17:00-17:15 Questions and Discussion
- 17:15-17:30 Edwin P.D.Pednault, Barry Rosen & Chidanand Apte:
The Importance of Estimation Errors in Cost-Sensitive Learning
- 17:30-17:45 Questions and Discussion
17:45-18:30 General Discussion, Closing Remarks