|
Artificial
Foreteller of Stocks and Commodities is only one of the examples
that demonstrates abilities of associative self-building neural
networks to solve very complex problems as prediction, classification,
repair, and diagnosis. To develop new systems that will process
any other type of data it is only need to change preprocessor
and GUI in system structure described for Artificial Foreteller.
Network Builder and Interpreter will remain the same. Below
there are several examples of the network implementation.
Mineral
and oil exploration
Preprocessor
receives on its entry presence or absence of a mineral occurrence,
geologic or formation characteristics and other relevant information
that describe known mineral deposits or oilfields. Based on
this information the system builds associative self-building
neural network. Then user enters into the system description
of area where there believe exist requested mineral deposit
or oilfield. Use of the system with proper trained associative
self-building neural network can greatly reduce mineral or oil
prospecting risks and of course reduce prospecting cost.
Solar
Flare Prognosis
Preprocessor
gets a lot of data pertaining to gamma-ray, neutron, energetic
charged particle emission, and other criteria that has been
observed from solar flares. During train session all known cases
with solar flares should be entered into Associative self-building
neural network. Then trained network can be easily used for
prediction future solar flares.
Tax
fraud catching
Preprocessor
should have judgmental benefits of an expert systems to prepare
tax data for input into Associative neural network. Files with
tax fraud cases that were found can be used as train data. During
prediction of tax fraud a queer report can be processed by associative
self-building neural network to make relevant decision.
Explosive
staff detection
Preprocessor
should get a set of data collected from different sensors that
describe situation with explosive staff detection. During train
session associative self-building neural network process available
data about such kind of cases and stores them in its structure.
At the same time it derives regularities and associations that
are common for known cases and stores them in its structure
too. During employment associative self-building neural network
compares current set of data collected from sensors that describes
situation with its experience and makes conclusion. In contrast
to human being and dogs the system is not tired of analysing
at this kind of staff.
As
it can be easily seen that current approaches can be used for
many different data domain.
Please
send any comments or suggestions by e-mail
| 