Neural Computing and Applications
Neural computing has become a commercially viable, and indeed very
successful technology within the last decade.
The concept is not new; it dates back to the 1950’s - although
supporting commercial software, hardware and expertise has facilitated
its massive growth over the last few years.
Furthermore, the DTI has recently supported a three year campaign
to promote neural computing as a valuable new technology in many
diverse areas of industry and commerce.
Neural computing represents a new approach to computing; operating
in a manner which is totally unlike conventional computing. The
method is inspired by the way in which a real neural network, the
human brain, works.
Neural computing therefore offers some abilities that characterise
the human brain - in particular:
- learning
- fast recognition of images or patterns - such as faces, sounds
and handwriting.
- making intelligent predictions and classifications based upon
prior information and learning.
Such capability is not available from conventional computing; at
least not without prohitory time and cost overheads.
At present, Neural computing is mostly developed using software,
with only a few hardware solutions being developed by organisations
such as the University of Hertfordshire.
The benefits introduced above make neural computing suitable and
valuable in situations where decision making and other ‘intelligent’
activity is required using a number of key data values.
It is often difficult to recognise the patterns and trends present
in data in order to extract the really useful information —
without using powerful techniques such as neural computing.
Case Study
A major tour operator put their extensive database to use - using
neural computing technology - to predict which holidays particular
clients were likely to go for. This information enabled then to
send the right type of holiday brochure to the right client.
This not only saved considerable expenditure in reducing the amount
of unnecessary literature but it also significantly increased the
rate of return on their direct mail operation.
Successful Application Areas
In addition to applications in targeted marketing, neural computing
can, and has, been successfully applied to a wide range of other
problems:
- Sales forecasting
- Customer profiling
- Financial forecasting
- Wine testing
- Medical test results analysis
- Credit scoring
- Recognition of speech, signature, character, handwriting and
fingerprints.
- Quality control
- Process control and optimisation
- Automated forms processing
There are many different ways in which neural computing can help
your business. For instance, credit scoring applications are used
to minimize the risk of lending to customers identified as potential
bad risks.
Fingerprint recognition, as with most forms of recognition, is
a case where neural computing represents a unique solution to a
problem which cannot be solve by any other means.
The aim of direct mail targeting applications is to reduce the
high costs of bulk mailing by targeting those customers that are
most likely to respond.
The main advantage of neural computing over other technologies
and techniques is the minimal length of time required to develop
an applications. Neural computing, because of how it works, is often
many times faster and is more powerful than other techniques.
Statistical analysis techniques, for example, are usually out-performed
by neural computing in terms of speed and accuracy.
Neural computing would not be so widely used if it were simply
an alternative to existing, conventional methods. It is therefore
found to be used in many different applications, where its special
capabilities make neural computing advantageous over other approaches.
The benefits are often clearly seen and relatively easy to calculate
— both in terms of potential cost savings and in greater performance
or rates of return. In addition, where an existing system is in
operation — using a statistical model, for example, direct
comparisons of performance and accuracy can easily be made.
|
