Transcript Presentation

Supporting harvest prediction
using
artificial intelligence techniques
Jonathan St Clair
Computer Science Honours
2003
Jonathan St Clair
STCJON003
[email protected]
10th September 2003
Background
Jonathan
St Clair
STCJON003
[email protected]
10th September 2003
Overview


On going research done to better predict harvest
figures
Often historical data is incomplete thus making
prediction difficult
Jonathan
St Clair
STCJON003
[email protected]
10th September 2003
Complex Adaptive Systems

Too many variables for management to optimise production for both
short and long term production

Not possible for management to work through every possible
scenario

Seasonal variations difficult to predict
Jonathan
St Clair
STCJON003
[email protected]
10th September 2003
Objectives

To identify aspects which could be meaningfully enhanced by the use
of AI techniques

To select the most promising opportunity within the prediction and
planning of the farm and



Jonathan
Describe the environment and its challenges in detail.
Select the most appropriate AI technique/s and describe their
application to the problem.
Illustrate how the farm management will benefit from this
application of technology to the business.
St Clair
STCJON003
[email protected]
10th September 2003
Deliverables

Interim report describing area of application for AI (I&J)

Software design document (UCT & I&J)

Final report (UCT & I&J)

Software prototype (UCT & I&J)
Jonathan
St Clair
STCJON003
[email protected]
10th September 2003
Impact

Enable management to quickly and reliably assess the impact of
changing any of a number of variables

Increase the ability of the farm management to prepare themselves to
meet a particular demand in the best possible way
Jonathan
St Clair
STCJON003
[email protected]
10th September 2003
Success Factors

The software must be shown to endorse or contradict decisions made
using the current management system

A number of test cases, of the farmers choosing, will be constructed
to allow for the farmer to make judgements in the normal fashion

The AI system will be tested on the same cases and if it is shown that
the system is consistently more correct, the system will be deemed
successful
Jonathan
St Clair
STCJON003
[email protected]
10th September 2003
Related Work

Robert M. Dorazio and Fred A. Johnson, Bayesian and Decision Theory – A
Coherent Framework for Decision Making in Natural Resource Management.

Andrew Wilson, Consumer Demand and the Future of the Supply Chain

Anet Potgieter, “Complex Adaptive Systems, Emergence and Engineering:
The Basics.”

Anet Potgieter, “Bayesian Behaviour Networks as Hyperstructures”

Fred Johnson & Ken Williams, “Protocol and Practice in the Adaptive
Management of Waterfowl Harvests”, http://www.consecol.org/vol3/iss1/art8/

Nils J. Nilsson, “Artificial Intelligence: A new Synthesis”
ISBN 1-55860-535-5, 37 -55, 343 – 346
Jonathan
St Clair
STCJON003
[email protected]
10th September 2003