October 23, 1998

Stan Matwin
School of Information Technology and Engineering
University of Ottawa

Machine Learning - challenges from applications

In this talk, we will review some of the challenging research problems that we have identified in the course of a number of applications developed at the Text Analysis and Machine Learning Group at the University of Ottawa (TAMALE). Only some of these problems deal with the actual learning algorithms. Often, major difficulties arise in the area of labelling real-life data with class labels, in feature engineering, and in the evaluation of the results of learning.

In our presentation, we discuss solutions to some of these problems. Pragmatic approaches to some of the above challenges have been developed in the course of our work on an oil spill detection system, an aircraft component failure warning system, a text classification system, and a software engineering fault prediction system. We will emphasize that all these problems need future work, and that systematic solutions will be necessary if data mining is to become a well-founded computing technology.

Back to the TAMALE home page