Current 6 month DJ 30  Forecasts-- Attrasoft Dow 5
Evaluation of the Past Forecast
Stock Indices

Dow Jones Industrials
S&P 500
Russell 2000
S&P MidCap 400
Dow Jones Composite
NASDAQ Composite
NYSE Composite
Dow Jones Transports
Dow Jones Utilities

Money Rates

Prime Rate
Discount Rate
Fed Funds Rate
3 Month Treasuries
1 Year Treasury
5 Year Treasury
10 Year Treasury
30 Year Treasury
30 Year Mortgage Rate

General Economy

Gross Nat'l. Product
Gross Domestic Product
Real Gross National Product, Fixed 1992 Dollars
Consumer Price Index
Producer Price Index
Inflation Rate
Housing Starts
Gold Prices
Oil Prices

Peronal Incomes
Personal Savings
Personal Saving rate

Business Data 

NAPM (National association of Purchasing management) Composite Index
Industrial Production Index
Total Industry Capacity Utilization
Business Inventories
Retail Sales, SA
Total Privately-Owned Housing Starts

Monetary  Data 

M1 Money Stock, NSA
M1 Money Stock, SA
M2 Money Stock, NSA
M2 Money Stock, SA
M3 Money Stock, NSA

Exchange Rates
Yen/US Dollar
German Mark/US Dollar
British Pound/USD
Canadian Dollar/USD

Dow Jones 30 
Individual stocks

Attrasoft Predictor 2.6 Data

To see the data, click an item on the left.

Current "Attrasoft Dow 5" Forecast
History of "Attrasoft Dow 5"

More on our Predictor Software.
Online order, online delivery from Digital River
Online order, hard copy delivery.
Mail Order


Attrasoft produces "neural network" software to predict: 

  • stocks;
  • financial parameters; and
  • economic parameters 

based on a number of more fundamental economic inputs which represent the U. S. economy as a whole. 

Neural networks allow a computer to truly learn from a set of training data to allow pattern recognition and to formulate generalizations. The name "neural network" is used to describe the computer algorithms because they are based on the operation of a network of nerves, such as the human brain, and these computer networks learn much the same way a network of nerves learn. 

Because a computer is allowed find its own patterns and generalizations in the training set, it reaches its own conclusions and essentially eliminates or side-steps inherent limitations in man-made forecasts or models--the limiting nature of the theories and assumptions on which the models are based. The technique allows for an objective model to develop which is not biased by human preconceptions, creative embellishments, or cognitive limitations. 

Most of the training sets utilize the following economic parameters as the foundation for developing the models on which the forecasts are based: 

Interest Rates 



Stock Market 


  • Stock value 
  • PE-ratio (Earnings should not be used because of the discontinuity: it changes once every 3 months) 
  • 200-day Moving Averages 
  • 100-day Moving Averages 
  • Stock/(100-day Moving Averages) ratio 
  • Section Indicator (for Example, phil.semecond.index) 
  • PR ratio (Price/Revenue, revenue should not be used because of the discontinuity: it changes once every 3 months) 
  • ... 


  • Population
  • Sentiment index 
  • Which foot ball team wins 
  • ... 

An example of a data set for Intel stock is: 

  • 3-month US Treasury Bills 
  • 30-year US Treasury Bills 
  • Consumer Price Index 
  • GDP (Gross Domestic Product) Growth Rates 
  • Dow Jones Industrial Average monthly growth rate 
  • NASDAQ Composite monthly growth rate 
  • Intel Stock value monthly growth rate 
  • PE-ratio 

And, of course, the predictions from the models are considered against the historical movement of an index, indicator, or stock as a double check. If the model's predictions are truly unrealistic, they are reformulated and rerun. Of course, we do not try to censor the forecasts except in extreme cases, since this is returning the potential for human bias into the loop. 

Neural networks are hard at work in many different industries and services such as credit approval, voice recognition, weather forecasting, securities forecasting, and product design.