1. Standard Scheme used by Applications to Predict Stock Market Trends

There are numerous applications on the market which attempt to predict the behavior of, for example: AMEX, NYSE, and NASDAQ. They are all built on the same scheme:

The standard part to manage the stock portfolio; This 1 st step is usually based on a formal technical analysis. This analysis provides software users with a set of

u n i v e r s a l t o o l s for the initial filtering of stocks.

The non-standard part to implement the ideas of well known stock market gurus about how to trade successfully. These ensuing steps are unique to each trader’s methodology. Initially they may be formal, understandable (and formally repeatable!), then progress to a realm which is impossible to adequately describe, i.e. the steps become l e s s

f o r m a l.

Thus, the second, non-standard part is of interest here.

www.inns.org. one of the best methods of solving the non-standard part, i.e. problems of automatic recognition, or recognition of non-obvious stock market parameters, is with n e u r a l n e t w o r k s.

M o d e l i n g o f b e h a v i o r a l a n d b r a i n p r o c e s s e s. see introductory literature below

Neural networks are formed from hundreds or thousands of simulated neurons connected together in much the same way as the brains neurons. Just like our brain, neural networks learn from experience, not from programming. The behavior of a neural network is defined by the way its individual computing elements are connected and by the strength of those connections, or weights.

Neural networks therefore can be trained to find solutions, recognize patterns, classify data, and forecast future events. They run fast, are tolerant of imperfect data, and do not need formulas or rules. For example, a neural network can learn to predict next weeks Dow, detect flaws in concrete, or recommend the number of jurors to be called on a given day. (J. Lawrence, California Scientific Software)

Generally, one of the main advantages of neural networks is that you can train them to recognize something w i t h o u t d e s c r i b i n g how to do that. It is similar to human brain: For example you can instantly distinguish a man from a woman, without going through the thought process: This object has long hair AND the object wears a skirt AND the object uses makeup AND. . And nobody has ever given you such algorithm. When you were a child you simply trained yourself on examples.

The same applies to neural networks. If you design a neural network properly and then give it a sufficient number of good examples, it can be trained to recognize some non-obvious relations.

Examples :

How neural networks were used in Japan .

The Yamaichi Fuzzy Fund uses the so called neurofuzzy approach to make financial forecasts.

It handles 65 industries and a majority of the stocks listed on Nikkei Dow and contains approximately 800 fuzzy rules in its expert system. Rules are determined monthly by a group of experts and modified by senior business analysts as necessary. In addition, the neural network itself is used to teach the application using real historical trading situations. The application uses both the fuzzy expert system and the neural network to create statements like: The trading situation today is similar to this pattern, thus we need to do this, this and this.

The system was tested for two years, and its performance in terms of return and growth exceeded the Nikkei Average by over 20%. While in testing, the system recommended to sell 18 days before the Black Monday of 1987. The system went into commercial operations in 1988.

How neural networks were used to forecast yearly floods in Central Russia .

About necessity of flood forecasting please click here

So, early flood warnings are extremely important. Among the main characteristics of any given flood are the maximum water level and the day when the water level reaches this maximum.

We developed and executed this project for the Ministry of Civil Defense of the U.S.S.R. in the late 1980ies. We created and taught a neural network using weather and flood data in the region beginning with early 1930s. It was taught to predict water levels 7-15 days ahead of the last data available to the system.

We tested the system one Spring. It was asked to forecast the day on which the flood level would reach its peak. Furthermore, how high the water level would be.

Our forecast was very precise, and the error margin was minimal. A comparison of our results with actual water levels provided by the Regional Department for Civil Defense showed that our neural network had misjudged the actual maximum water level by merely 10 cm and was off only by only one day from the date the flood reached its peak.

N o b o d y h a s e v e r designed a software product that can recognize n o n -o b v i o u s stock market parameters better than a successful trader can. However, neural networks can be used for the complex filtering of stocks and thus substantially reduce a trader’s work.

A number of software products, which use neural networks to predict stock markets, already exist, e.g.

However, these are merely u n i v e r s a l tools and not really neural network applications or solutions which we offer. By analogy: They are the screwdrivers, hammers, nuts and bolts to assemble a car; not the operational vehicle.

Based on William J. ONeils Principles

We have developed a neural network that implements some ideas of William ONeil. He says that certain patterns on the stock charts are buy or sell signals, and the main problem is to recognize them.

Let me begin with the short description of the ONeils winning ideas and of how we implemented them. In short, ONeil insists that success awaits you on the following route:

1. Choose stocks meeting the requirements of the SmartSelect® Corporate Ratings — six proprietary and exclusive research ratings designed by William J. ONeil, Chairman & Founder, Investors Business Daily. They include: