"General Query Settings"
- "Class/Examined Field Selection" list - a list of
fields from the table, currently selected on the
Data Set Input page. Determine the
Class field by selecting a field from the list. It is recommended to
select fields, containing the most important data. For example, it could
be a field, that contains information that marks loyal or fraudulent
customers, important financial data.
- "Selection of Fields"
control - contains a
list of fields from the table, currently selected on the
Data Set Input page. The list doesn't
include fields of unsupported types and the field, selected as the
"Class/Examined Field". Fields, marked in the list, can be used for creating
Decision Rules and
Decision Trees. Note:
Marking the fields is necessary for
data mining (Rules or Decision
Trees creation). The Statistics can be obtained without marking the
fields.
After selecting the Class/Examined Field the program specifies the
selected
field's type. In case if the field is of text or integer type, the following
controls will be displayed:
- "Classes List"
list -
depending on the Class Field type it will contain a list of determined
classes. Select Classes to be analysed.
- "Total Number of Classes:" - displays the overall number
of Classes.

If the examined filed is of float type, the Classes list
control is replaced with controls for float Classes creation.


With the help of dispayed controls you can create a new class manually, by
puting it's minimum and maximum values into corresponding fields and pressing the "Add Class to
the List" button. You can also remove the selected class from a list,
by pressing the "Delete from the List" button or clear all classes
with the "Clear List" button.
Press "Create Classes" button after forming the list of classes.
This will save the list for further analysis.
Rules Query Settings
The "Rules Query Settings" group box contains the following controls:
- "Element belongs to class(%)" - input the
probability for Decision Rules.
- "Minimum number of cases for a Rule" - the rules,
describing less than minimum number of cases, will be ignored by the program.
- "Float fields minimum value" - determines
minimum float values that Estard Data Miner will use for the analysis.
- "Create full Rules" control - marking this control
will make Estard Data Miner to find all Decision
Rules, without cutting off unnecessary elements.
- "Create minimized Rules"
control - marking the
control will allow to create minimized Rules, enougth for Classes analysis.
The overall number of Rules might be less than with "Create full Rules"
option marked.
- "Maximum Rules number for a Class" -
this control allows to
limit the number of created Rules. For example, if setting controls value
to 10, the program will create no more than 10 Rules for each Class.
Decision Tree
Query Settings
The Decision Tree can be created in various ways.
- "Build Automatically" control- mark it if
you want the program to decide how to create the Decision Tree.
- "Build Tree for Each Field" control - mark it if you want
to recieve a Decision Tree for each Field, marked in the "Selection of Fields"
control. Time, necessary to obtain Decision Trees is proportional to number of Fields used in analysis
processing.
- "Build starting with Field" control - mark it to create a concrete Decision Tree, strarting with the Field selected
in the list below the control.
- "Build a Single Tree" - in case if some Classes were
not defined after building the first Tree, Estard Data Miner will try to build
a new Tree. Marking this control will make Estard Data Miner to create a single
Decision Tree.
- "Stop After defining All Classes" control - will stop
Decision Trees creation in case if all Classes were defined.
Decision Tree Output Settings:
- "Clear previous Trees before adding new ones" option - clears up the
displayed Decision Trees before adding
new Decision Trees.
- "Add new Decision Trees to previously created" - if
some Decision Trees are displayed in the control on the "Decision Tree
Output" page, they won't be removed when creating a new Tree.
|