MPhil Thesis Defence


"Mining High-Utility Plans from Plan Databases"

By

Miss Hong Cheng


Abstract

An important data source for data mining is the databases of plan
executions.  How to automatically find high-utility plans from the plan
Databases is an interesting problem.  Traditional data mining algorithms
focused on finding frequent sequences without utility consideration
through sequential mining.  Our contribution is a novel algorithm, which
automatically generates plans from large databases by combining data
mining and AI planning.  Our planning objective is to find high-utility
plans that convert groups of records from a less desirable state to more
desirable ones.  An improved sequence-mining algorithm combining frequency
and utility is employed to focus on an abstract space of the original
problem in which a frequent subset of plans with high utilities forms the
basis for an approximate planning algorithm.  Our formulation and solution
avoid both the shortcomings of traditional AI planning, which relies on
the precise knowledge of actions' logical formulations, and the
computational problem faced by many MDP formulations in uncertainty
planning.  We show through empirical results that planning using our
combined algorithm produces high-utility plans that are close to optimal
quality while at the same time, is much more efficient and scalable than
planning using MDP and exhaustive search.


Date:			Thursday, 3 July 2003

Time:			10:00a.m.-12:00noon

Venue:			Room 2302
			Lifts 17-18

Committee Members:		Dr. Qiang Yang (Supervisor)
			Dr. Dit-Yan Yeung (Chairman)
			Dr. James Kwok


**** ALL are Welcome ****