Home' Know How : KnowHow Issue 1 Contents 8
in the news
A company servicing the mining industry that can trace its success
to a simple innovation has been awarded the prestigious 2013
Australian Museum Rio Tinto Eureka Prize for Commercialisation of
Researchers from Curtin University and the CRC for Spatial
Information (CRCSI) spotted a need to undertake more rapid
testing of the machinery used to crush and grind rocks to extract
minerals, which used to be tested by personnel undertaking laborious
measurements. The machines use around 60% of a mine’s energy
and can cost up to $100,000 to replace.
The researchers came up with a 3D scanning laser system that
measures 10 million points or more in just 30 minutes, so rapidly
testing the machinery for faults. The technology was brought to
market by a company formed for the purpose, Scanalyse, which was
later bought by Finnish mining giant Outotec.
The company is a good example of how to move innovation from
R&D to industry, said Peter Clarke, Scanalyse’s CEO. The company
is now set up in South Africa, Chile and the U.S . and runs projects
across 20 countries. “It drives efficiency through the mining system,”
said Graeme Kernich, CRCSI’s deputy CEO. “It’s a good business.
It’s going places.”
– Heather Catchpole
Mining innovation recognised
Predicting the accuracy of
daily stock returns is big
business to finance. Now,
researchers have found these
returns can be improved by
mining the information from text
sources – such as news reports and
online chat forums – as well as
from quantitative financial data.
The method has the potential
to give traders an edge in the
stock market or, if widely
established, even improve the
informational efficiency of
the market itself, according to
researchers. Zhendong (Tony)
Zhao, a PhD student at the
Capital Markets Cooperative
Research Centre (CMCRC),
said in comparison to more
models, their system achieved
an improvement of some 3%
on other methods of predicting
stock price movements.
The comparative study assessed
the accuracy of four different
models with different weightings
of text and quantitative data from
2010. This included documents
such as financial reviews and
news announcements from the
Australian Securities Exchange
(ASX) as well as figures such as
a company’s past daily returns and
capital size, taken from Reuters.
Text mining is about
transforming unstructured text
into structured quantifiable data
that can be analysed by a computer.
The process involves identifying
patterns in text so that machines
can automatically extract relevant
and interesting information.
Researchers analysed the text
using a bag-of-words method,
which assessed the frequency of
particular terms in a document,
representing the result as
a mathematical vector.
“For each [news]
announcement, we combine
this vector with corresponding
quantitative numbers to predict
the magnitude and direction of
price movements,” said Zhao.
Matthew Clifton, a
quantitative developer and
researcher at CMCRC, who is
involved in the research, said
the next step is to evaluate the
predictive model against a live
trading market using various
back-testing tools. “And if that
checks out, it may be used in
a live context to trade,” he told
Zhao and his colleagues will
soon present their findings to
a computer science conference
before publishing their results.
Clifton said their hope is that
researchers in both computer
science and finance will explore
this area further.
“In finance it’s kind of
the Holy Grail to be able to
incorporate extra information
in stock prices. Zhao is bridging
the gap between computer
science and finance in that
regard,” Clifton explained.
John Quiggin from the School
of Economics at the University of
Queensland, who wasn’t involved
in the research, commented that
the standard view in economics is
that profitable prediction of stock
market movements on the basis of
publicly available information is
impossible. But he also explained
that long-term changes affecting
entire markets have sometimes
been the exception, and that if
Zhao’s study was established, it
“would have major implications
for financial economics”.
Text mining helps analysts take stock
The researchers believe the
method has the potential to
revolutionise predictions of daily
returns, as well as have a significant
impact on the stock market.
“ We’ve been working on this for
years, and Zhao has shown some
good results,” added Clifton.
class measuring and
modelling tools have
saved the mining
Links Archive KnowHow Issue 2 Navigation Previous Page Next Page