Knowledge Engineering a Bayesian Network for an Ecological Risk Assessment (KEBN-ERA)

Navigation:

 

Introduction

Background
- Knowledge Engineering Bayesian Networks  (KEBN)
- Ecological Risk Assessment (ERA)

Events
- Phase 1
- Phase 2
- Phase 3

Conclusions

Downloads

Links

References

Honours project web-page for Owen Grant Woodberry.

My 2003 honours project: Knowledge Engineering Bayesian Networks for an Ecological Risk Assessment, is the focus of this web site. My supervisors are Dr Ann Nicholson, Dr Kevin Korb and Dr Carmel Pollino.

You can contact me in regards to anything on this site at: Owen.Woodberry@csse.monash.edu.au.

Thesis Abstract:

This thesis develops upon existing research in the field of knowledge engineering of Bayesian Networks (KEBN), specific for a complex application domain.  The application studied in this project is an Ecological Risk Assessment (ERA), which is being conducted in the Goulburn Broken Catchment, Victoria, Australia.  The objective of this ERA is to model the effects of various management interventions on native fish populations.  The system being modeled is complex, with multiple variables and interactions.  Relationships between variables within the system are poorly understood and data is limited.  Although the knowledge engineering of BNs can be an exceptionally difficult task when developing complex models, their representative power justifies their use. 

To assist in the building of complex BNs, formal knowledge engineering techniques and tools are required.  This thesis used the ERA model to identify existing and potential techniques and tools that are useful for development of the model ontological, qualitative and quantitative components.  Tasks of elicitation, evaluation and implementation using both domain expert and automated methods were investigated and techniques/tools were applied or developed where none existed.  The results of these investigations were an increased representational power and validity of the model.  In addition to the development of tools and methodologies for evaluating and improving the model, constructive evaluation suggestions for future model development were identified.