home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Trading on the Edge
/
Trading On The Edge - CD-ROM Toolkit (Wayzata Technology)(2031)(1994).bin
/
pc
/
pc_files
/
venddemo
/
visanlst
/
t001
< prev
next >
Wrap
Text File
|
1993-01-27
|
6KB
|
150 lines
.LsnName Information Harvesting
.LsnDesc Flow diagrams illustrating the functions of the
Information Harvesting software.
.LsnScts What is IH?, Rule Generator, Inference Engine, Advantages, How it works
.LsnPrms Estimation data glosintr
.What
~G~m 40 40 0 0
Finding relationships in data~m 100 100 6 6
Classification
Prediction
Understanding~m 40 40 16 16
Exploiting database information
^!H` Representing how systems work !@!t 200 120 600 320
~c~w~m 0 20 8 0^!`Representing!: corresponding to in essence. ~
Here, a computer model that simulates the behavior of a system.
^!`System!: a regularly interacting or interdependent group of ~
items forming a unified whole, ~
i.e. any business, organization, governing body, country, economy or ~
technology.
^!`Work!: to function or operate.`~
.Rule ~c~
~C00~C00
~C00Prediction~C00
~C00Variable~C00
~C00~C00 ç ç
~C00~C00 ê ê
~C00Numbers~C00 àå Rule
~C00~C00 àå Induction àå Visual
~C00Categories~C00 àå Data Analyst
~C00~C00 ç
~C00Times~C00 àå Prep ê
~C00~C00 Rules Resolution
~C00Opinions~C00 àå & Importance
~C00~C00 ç Distributions
~C00Feeds~C00 àå Binning ê Quality
~C00~C00 àå Inference àå
~C00Money~C00 àå Engine
~C00~C00
.Inference ~c~
~C00~C00
~C00~C00 Rules
~C00~C00
~C00~C00 ç
~C00~C00 ê
~C00Numbers~C00 àå
~C00~C00 àå
~C00Categories~C00 àå
~C00~C00
~C00Times~C00 àå Data Inference
~C00~C00 àå Predictions
~C00Opinions~C00 àå Binning Engine
~C00~C00
~C00Feeds~C00 àå
~C00~C00 àå
~C00Money~C00 àå
~C00~C00
.Advantages
~G~m 20 20 0 0
^!H` Quantitative learning from experience !@!t 200 235 600 400 ~m 0 0 8 0~
~c~wPeople learn from experience. ~
This is the most effective way to learn.
Information Harvesting learns from the experiences captured in a database.
The Visual Analyst provides the tools for you to understand ~
what Information Harvesting has learned.` ~m 20 20 16 16
^!H` Flexibility !@!t 200 250 600 400
~c~wFlexibility~m 0 0 8 0
Rules can adapt to multiple driving mechanisms since they do not assume ~
that a single global model will explain the data. ~
All other modeling technologies, except Expert Systems, assume a ~
single global model.`
^!H` Industrial Strength !@!t 200 250 600 400
~c~wIndustrial Strength~m 0 0 8 0
The technology can handle hundreds of variables and millions of data rows.
Rule generation is robust in the face of missing values, erratic values ~
and nonlinear behavior.`
^!H` Fit Consistency !@!t 220 240 600 400
~c~wFit Consistency~m 0 0 8 0
The rule generator uses a technique similar to statistical jackknifing ~
to enhance the consistency of predictions by actively discriminating ~
against overfitting. ~
You can trust the rules to work as well on new data as on the design data.`
^!H` Data ranking !@!t 200 140 600 300
~c~wData Ranking~m 0 0 8 0
The rules assign a ^!`quality! to the information content of each row ~
of data. ~
This provides the information you need to evaluate each individual ~
prediction as well as to upgrade the quality of your data acquisition.`
^!H` Understanding !@!t 200 120 600 330
~c~wUnderstanding using the Visual Analyst:~m 20 60 8 0
Global accuracy of the rules
Importance of each variable for providing information about the ~
prediction
Cross relationships between all possible pairs of variables
Data ^!`information quality! and its influence on every aspect of the results`~
.How
Suppose we want to predict the temperature based on the weather.
Consider the following table:
~b~m 80 80 0 0
Sunny Hot
1 1
1 1
1 1
1 0
1 0
1 0
1 0
1 0
0 1
0 0
0 0
0 0
~t~m 0 0 8 0
\We can make an accurate prediction that the temperature will be hot
only 3 out of 8 times.
This is not a great predictor.
\\Suppose we add further information conerning the season:
~b~m 80 80 0 0
Sunny Summer Hot
1 1 1
1 1 1
1 1 1
1 1 0
1 0 1
1 0 1
1 0 0
1 0 0
0 0 1
0 0 0
0 0 0
0 0 0
~t~m 0 0 8 0
\If the weather is sunny and the season is summer,
we can predict it will be hot 3 out of 4 times.
\\\Information Harvesting finds its rules by considering massive
numbers of such truth tables.