TITLE: Classification by Conditional Probability Estimation

PRESENTER: Professor Geoff Webb    http://www.cm.deakin.edu.au/webb

AFFILIATION: School of Computing and Mathematics, Deakin University

DATE: Friday 12 April 2002

TIME: 12:00 noon - 1:00pm

PLACE: Seminar Room K17

ABSTRACT:

The naive Bayes classifier is a computationally efficient, elegant, and theoretically well-motivated approach to classification learning. Despite its simplicity, it has high classification accuracy, especially for small data sets. This talk presents our research into improving naive Bayes, retaining its efficiency, elegance, and direct theoretical base, while further strengthening its classification accuracy. I present two key techniques, one that delivers very high classification accuracy at some computational cost and the other that delivers substantially improved classification accuracy at modest computational cost.

BIOGRAPHY OF SPEAKER:

Geoff Webb holds a Personal Chair in the School of Computing and Mathematics at Deakin University. He has published over 90 papers in the areas of machine learning, data mining, and user modelling. He has received more than $840,000 in national competitive research grants. He is a member of the editorial boards of User Modelling and User-Adapted Interaction (UMUAI) and Knowledge and Information Systems (KAIS).

Seminar Convenor:

Bernhard Hengst
Tel: +61-2-9385-3988,
E-mail: bernhardh@cse.unsw.edu.au
School of Computer Science & Engineering, UNSW.