<<Back            Contents            Next>>

Digging deep

Millions of dollars in untapped revenue sit in the vast electronic databases of corporations worldwide. DEREK BROWN reports

Monash researchers at the university’s School of Business Systems are helping companies around the world access the wealth of information contained in corporate databases using an innovative technique called data mining.

In data mining, mathematical models called neural networks are used to sort through data to identify patterns. Organisations can apply these patterns in a variety of ways – from setting insurance premiums to working out what conditions encourage punters to bet on a horse, says Dr Kate Smith, director of the Data Mining Research Group in the Faculty of Information Technology.

"Neural networks are inspired by the human brain and the way humans learn through repetition and example. The networks sort through thousands of examples of customer data to find patterns," Dr Smith says.

"AAMI insurance, for example, came to us because they wanted a better and fairer way to work out premiums."

The Data Mining Research Group used more than 30,000 entries of customer transactions recorded in AAMI’s database to work out what characterised good and bad driver behaviour – information which the company could use to set fairer premiums.

And industry has been happy with the results. AAMI's manager of research and development, Mr Malcolm Brooks, says working with Monash's School of Business Systems provided concrete results that could benefit the company.

"The School of Business Systems had a real feel for something we knew little about. We took on board all the directions they suggested," Mr Brooks says.

Dr Smith says the New South Wales TAB (gambling agency) used datamining to  identify betting patterns: "We looked at factors such as track conditions, weather, the number of horses running and the odds placed on the favourite in each race and how they influenced betting."

Information on the different factors was fed into the neural network, uncovering patterns in betting behaviour. Dr Smith was able to use the results to create a more accurate prediction on the level of betting on any particular day.

"We were able to give the TAB the conditions affecting betting levels. Some rules – for example if the track is known to be wet, then turnover is up 10 per cent – could prove very useful to the TAB," she says.

As an independent source of information on data mining in Australia, Dr Smith and the Data Mining Research Group are highly sought after by Australian businesses and organisations.

"Apart from our group, the only other place you can get information on data mining is through commercial vendors pushing particular software packages," she says.

"As well as AAMI insurance and the TAB, we are working with a number of Australian businesses, including several major banks and financial institutions interested in looking at loans and approving credit, and market research firms looking at identifying and targeting particular markets."

Dr Smith believes the growing popularity of data mining can be traced to several factors, including more companies maintaining data warehouses, increasing holding capacities, faster computers and sophisticated data sorting methods. "Analysing this data is now much easier as a result," she says.

Even though the technology and expertise exists, Dr Smith believes that most Australian businesses, while interested in the process, are still reluctant to make a big investment in data mining.

"This technique is very popular in the US and there are many success stories of companies being transformed through data mining," she says. "Many Australian businesses, while interested and willing to pay for an original consultation, are still too conservative to take it to the next step."

But Dr Smith believes increased use of data mining in Australia is just a matter of time.

ACTION: The Data Mining Research Group provides independent advice on data mining, conducts corporate training, and is a source of qualified graduates. Visit www.bs.monash.edu.au/staff/KATE/datamining.html

 <<Back            Contents            Next>>