Room A529,
Wednesday 26th November 1997 (6:30pm)
Prof. Max Bramer (University of Portsmouth)
DATA MINING OF RULES FROM EXAMPLES: VISION OR MIRAGE?
ABSTRACT
In recent years the considerable commercial potential of the subfield of Machine Learning known as Data Mining has increasingly become recognised. One of the key technologies of data mining is the automatic induction of rules from examples, particularly the induction of classification rules. There are two principal motivations for work of this kind:
(a) the need to analyse the ever-growing volume of data held by many organisations, where the aim is to discover hidden relationships in the data
(b) the well-documented 'knowledge acquisition bottleneck' in the development of rule-based expert systems, where the supplying of significant case examples to an inductive learning program has frequently been advocated as a more natural means of knowledge transfer than the conventional techniques of Knowledge Engineering.
Although some of the world's largest expert systems have been constructed using rule induction techniques, there are a number of significant technical problems to overcome.
In this talk I shall concentrate particularly on the well-known ID3 family of rule induction systems. I shall seek to illustrate the potential value of the rule induction approach, highlight some of its major weaknesses and suggest some possible ways in which these might be overcome. No prior knowledge of the subject will be assumed.
Max Bramer is Digital Professor of Information Technology at the University of Portsmouth, currently on sabbatical leave at Birkbeck College, London. He is chairman of the British Computer Society Specialist Group on Expert Systems (SGES) and a member of the Artificial Intelligence Professional Group of the IEE. He is Chairman of the organising committee for the SGES conference Expert Systems '97 in Cambridge in December.
Professor Bramer has been actively involved in research in the field of knowledge discovery for over ten
years.
He supervised the successful PRISM and CUPID rule induction projects and currently leads the University of Portsmouth's
Artificial Intelligence Group which is conducting research into both symbolic and subsymbolic techniques of knowledge
discovery (neural networks, genetic algorithms, case-based reasoning and rule induction).
His principal current interest is in bringing domain knowledge to bear on the discovery process.
LOCATION AND TRAVEL DETAILS
Room A529:
Use City University's St John Street entrance, opp. Whiskin St EC1. The very large clock hanging over this entrance
makes it easy to spot. Obtain a visitor's pass from security. Use the corridor leading away from the street. Take
the lift on the left from level 2 to level 5. Exiting the lift, turn right, then all corridors take you to A529.
TRAVEL
From ANGEL tube station: Turn left on exit. At the crossroads go straight ahead into St John Street. Carry on for
c. 500 yards (8 min). (Angel is one stop on the Northern Line City Branch from Kings Cross.)
From Waterloo: A taxi or no. 4 bus (to Percival St) beats the tube.
The Evening Lectures are free to both members and non-members of SGES.
For further information contact:
Ilesh Dattani dattani@compsci.bristol.ac.uk
Tel: +44 (0)117 954-5160
or
David Dodson (local arrangements at City University) dcd@cs.city.ac.uk
[Evening Lectures Home Page] [SGES Home Page]