1. Introduction

1.1 Why Predictor

1.2 Why Attrasoft Predictor

1.3 How to Operate Attrasoft Predictor

1.4 Phase 1: Data Set and Data Collection

1.4.1 Data Set
1.4.2 Data Collection
1.4.3 Preprocessing
1.5 Phase 2: Operate Predictor

1.6 Predictor's Sister

1.7 Predictor's Version


Why Predictor

Making decisions based on your projection of the future events is a way of life.

Assume you run a coffee shop and your raw material is coffee. If you project the coffee price is going significantly higher, then you probably will keep the coffee inventory as high as possible. If you project the coffee price is going significantly lower, then you probably will keep the coffee inventory as low as possible. How do you make this kind of projection? Most likely you will make a prediction based on your historical experience.

If your decision is whether to make an investment to build a $100 million factory, you'd better be sure that your projection will be right. In particular, you should keep your emotions out of the decision process.

In general, a decision is made based on several factors. Some of them can be measured and have a historical track record. Many corporations already have their database in place that contains years of historical data. It is often overwhelming because of the sheer volume to mine the data for the purpose of strategic thinking.

This is where software can help you to predict these factors based on your historical database. The software can learn years of experience from your database and use the knowledge to tell you what will happen if this were in the past. This provides you with a baseline for your decision making process. The software can learn years of experiences in seconds and is ready to serve you as if it is one of your staff.

Why Attrasoft Predictor

Attrasoft Predictor is a fast terabyte data processing tool for your database. It uses a sequence of numbers to predict the next row of numbers in line. It uses historical data to predict future possibilities. It does not matter what you want to predict.

90% of the work to use the Predictor is to prepare your database containing your historical data. (Many corporations already have their database in place that contains years of historical data.) The remaining 10% of the work is to operate the Predictor (two clicks). The Predictor is especially good if you have a terabyte or gigabyte database because of its accuracy and speed.

The advantages of the Attrasoft Predictor are:



Figure 1. Attrasoft Predictor.

Once your data is prepared correctly, the Predictor is able to provide you with rated predictions on any subject or any problem. Attrasoft Predictor can:

How to Operate Attrasoft Predictor

The Attrasoft Predictor makes patterns out of complicated problems. The Predictor is based on the neural network technology developed at Attrasoft. A neural net learns from past experience; this makes the neural software operation different from other software: you have to train the software first. You have to teach the Predictor by showing it your historical data.

The operation has two phases:

  1. Your raw data -- 90% of the work
  1. Using the Predictor -- 10% of the work

In section 1, we discussed two examples that require the projection of the near future. To show you how to use the Predictor, we, however, will not go into "very complicated" problems like a business decision of whether to build a factory or not. We will concentrate on "very simple" problems where the fruit of the prediction is immediate: like the stock prediction. The decisions are very simple:


Phase 1: Data Set and Data Collection

If you want the software for predictions, chances are you already have decided the "data set" and have collected volumes of data. This will make using the Predictor very simple.

Data Set

Let us start from the beginning:

a "data set" is a set of relevant factors related to your prediction problem.

Throughout this menu, we will use the stock market as an example.

What are the factors for the stock market? i.e. what is a data set for a stock? There are hundreds of them:

An example of a data set for Intel stock is:

Data Collection

Once a data set is chosen, the next task is to collect data. For the stock market, data can be collected free from the Internet. We will postpone this topic to chapter 5.


Sometimes, the data has noise (statistical fluctuation). The noise can be removed by preprocessing; for example, using the 100-day moving average of a stock instead of the stock directly. Calculating the moving average of a stock from the stock data can be done within seconds with only a few clicks. Again, we will postpone detailed discussion of this topic to chapter 5.

Phase 2: Operate Predictor

The software is a black box. The black box only has 3 parameters to setup:

The user's job is mainly in phase 1: set up a data set and collect data, if necessary, preprocess data. Phase 1 takes 90% of the work. Phase 1 is simply getting your historical data together into one file.

Phase 2, operating the Predictor, has three steps:

  1. Prepare input file (put all your historical data into 1 file);
  2. Link the file to the black box; and
  3. Run.

Once your historical data is all in one file (step 1), this file is linked into the black box for training (step 2). In English, this means the Predictor is learning your problem, experience, and values of your problem, starting from an empty but very smart brain. After that, the neural network black box is ready to make a prediction by clicking one command (step 3).

The Attrasoft neural network is the based on the so-called Boltzmann machine neural network; this special type of neural network will predict a distribution, which is a set of rated possibilities. (This is similar to the Internet search engine: when you use a Internet search engine, you will get a set of rated possibilities.)

Predictor's Sister

The Predictor predicts downward into the future. If you want to predict to the right, you need to use the Predictor's sister: DecisionMaker.

Attrasoft DecisionMaker can:

Predictor's Version

The Predictor has several versions:

Version                 Neurons             prices

Standard version     10,000             $99.99

100K version         100,000             $499

250K version         250,000             $999

1M version             1,000,000          $9999

Other customized versions                 $999 or up

Consulting fee:

The free customer support covers how to operate your Predictor.

If you need help on how to encode your problem, how to choose a data set, how to convert string data to numerical data, how to handle missing data, ..., the consulting fees are:

  • $250 per problem if the problem has 1000 rows or less;
  • $500 per problem if the problem has 1,001-10,000 rows or less;
  • $1000 per problem if the problem has more than 10,000 rows.