Chapter 6 • Managerial Support Systems 245the final decision. Neural networks are being used to
predict the probability of bankruptcy to help banks make
lending decisions.
Neural networks are also being used in investment
and trading applications. In some cases major companies
are using neural networks to manage their pension fund
portfolios. Another application detects common character-
istics among stocks to determine whether a stock is on the
verge of a breakout. Neural networks are being used to
predict the next day’s closing prices of stocks and to group
mutual funds based on performance measures.
Another use of neural networks is in targeted
marketing, where marketing campaigns are targeted to
potential customers who have the same attributes that
resulted in sales for previous campaigns. Spiegel Brands,
Inc., which depends on catalogs to generate sales for its
mail-order business, uses a neural network as a way of
pruning its mailing list to eliminate those who are unlikely
to order from Spiegel again.
Neural networks are also being used to improve
security. A security system has been developed that uses
neural technology to recognize a person’s face to grant that
person access to a secured area. A computer network
intrusion protection system, based on a neural network,
conducts a real-time assessment of each visitor to a net-
work, and if it notes behavior that indicates an attempted
security breach, it automatically terminates the intruder’s
access (Orzech, 2002).
In the late 1980s and 1990s, expert system and
neural network applications received a great deal of hype
in the popular press. The AI applications were supposedly
going to solve many of the decision problems faced by
managers. Today, industry has adopted a more realistic
view of AI applications: AI is not a panacea, but there are
a significant number of potentially valuable applications
for AI techniques. Each potential application must be
carefully evaluated. The result of these careful evaluations
has been a steady growth, but not an explosion, in the
development and use of expert systems and neural net-
works to help businesses cope with problem situations and
make better and more consistent decisions.Virtual Reality
Virtual reality is a fascinating application area with rapidly
growing importance. Virtual reality (VR)refers to the use
of computer-based systems to create an environment that
seems real to one or more senses (usually including sight)Neural Networks in Medicine
Neural networks have become increasingly important in the field of medicine in areas such as drug
development, patient diagnosis, and image analysis. Two major applications of neural networks occur in
the detection of coronary heart disease and the processing of electroencephalogram (EEG) signals.
Identifying specific characteristics in medical imagery is a type of image processing problem, and
neural networks can be used to handle such problems. For example, a clinical study has shown that neural
networks can provide a useful tool to aid radiologists in the mammography decision-making task. When
lesions are found in a mammogram, the performance of a neural network in distinguishing between
benign and malignant lesions has been found to be better than the average performance of a radiologist
without the aid of a neural network. The conclusion is obvious: The best performance will be achieved
when the radiologist has the assistance of a neural network.
At Anderson Memorial Hospital in South Carolina, neural networks embedded in a hospital
information and patient prediction system have improved the quality of care, reduced the death rate, and
saved millions of dollars in resources. Using California Scientific’s BrainMaker software, a separate neural
network has been trained (developed) for each of 473 primary diagnoses to enable the hospital to classify
and predict the severity of illness and the use of hospital resources so that quality and costs issues can be
addressed fairly. A neural network has also been used to predict the mode of discharge—from routine
discharge through death—for each diagnosis. Based on the resulting predictions, expenses at the hospital
have been reduced by fewer unnecessary tests and procedures, lowered lengths of stays, and other
procedural changes. For a given diagnosis, about 400 to 1,000 cases were used for training the neural
network, with length of stay in the hospital as the primary variable to be predicted based on 26 input
variables such as the number of body systems involved (e.g., cardiac and respiratory), number of compli-
cations, smoker or not, diabetic or not, age, sex, race, marital status, and number of previous admissions.
[Based on California Scientific Software, 2010; NeuroDimension, 2010; and NeuroXL, 2010]