MDMC Research Areas
Our members are actively researching all of the following areas and
techniques, as well as others.
- Artificial Neural Networks (ANNs)
- ANNs are a proven technology for performing
non-linear regressions. We use neural net packages in SAS,
NeuroShell, Matlab and other software in consulting
activities.
- Bayesian Networks
- Bayesian networks have been developed over the past 15 years as
a sophisticated technology for representing and reasoning about
uncertainty, performing decision analysis and planning under
uncertainty, and causal modeling. We work with a variety of
commercial and non-commercial BN tools and have developed CaMML
(Causal MML) for the automated learning of Bayesian nets.
- Bioinformatics
- Extracting information from genetic data such as DNA and
protein sequences.
- Classification
- Various types of data classification, including medical
data such as MRI images.
- Data Visualisation
- Graphical representation of data analysis results and
processes.
- Decision Trees and Decision Graphs
- Decision trees and their more efficient generalization,
decision graphs, perform supervised learning for
classification. Because of their clear structure and easy
interpretation, they are very useful when you are trying to learn
simple rules or suspect that your data has a hierarchical
structure. They can also be used for general function
approximation (like neural networks), and are often embedded in
other models (like Bayesian networks) to find efficient
represenations of complex structures. Our own decision tree and
decision graph software has repeatedly outperformed commercial
packages, often by large margins.
- Econometrics
- A branch of economic modeling and forecasting which may
incorporate several methods listed here, and otherwise.
- Effects on the GNP of government funding of education versus
more general spending such as tax cuts
- Testing the efficient markets hypothesis
- Effective data warehousing and data mining
- Knowledge Engineering
- Sometimes you don't have a large database to search, but rather a few human experts who have learned the domain quite well. Knowledge Engineering is the art and science of putting expert knowledge into computer systems. This is most often done in medicine, where the most successful systems use Bayesian networks.
- Image Processing and Pattern Recognition
- Neural networks for diagnosis, understanding, and treatment of
various medical conditions.
- Intelligent decision support
- Business assement for financial services.
- Intelligent tutoring
- Using Baysian networks to infer student misconceptions about
various areas of study.
- Machine learning
- Most of data mining falls under the academic category of
Machine Learning, which is part of Artificial Intelligence
generally. The members of MDMC have developed some of the most
successful Machine Learning algorithms in the world, in many of
the specific domains listed on this page. We also analyse and
write about improving methods for evaluating and comparing machine
learners, and have written tools to use these methods.
- Market response modelling
- Useful for predicting consumer responses to new or existing
products and advertising techniques.
- Minimum Message Length Induction
- A rigorous information-theoretic method for performing
statistical inference, or evaluating the performance of various
models. Robust and scale-invariant, MML methods will fit the data
but not the noise. MML was invented by Monash Computer Science's
Foundation Professor Chris Wallace. MDMC develops the basic theory
of MML, invents alternative numerical approximations to strict
MML, and applies MML to various fields using many kinds of
models. For more information see Lloyd
Allison's MML web page and/or David
Dowe's MML web page.
- Natural Language processing
- Argument generation and analysis.
- Neural Networks
- See Artificial Neural Networks above.
- Pattern analysis and behaviour prediction.
- Weather prediction
- Predicting missing person behaviour.
- Planning
- Planning is the general problem of considering alternative
ways to get from here to there, and choosing the best. It may be
path selection, or experiment design. We have used it for
automated aircraft navigation, among other things.
- Resource Optimisation
- Optimal resource allocation for search & rescue
- Statistics
- Almost all Machine Learning techniques rely heavily on
statistical inference. Consequently, MDMC has a great deal of
expertise and can offer short courses on statistics, statistical
inference, and modeling. We specialize in Bayesian methods
- User modelling
- Predicting user's requests to pre-send documents on the WWW,
also useful for database queries.
|
