Quick
Definitions
|
|
|
| Data Mining |
|
Data Mining
is the automatic extraction of potentially useful
information from raw data. |
| |
|
|
| Rule Induction |
|
Rule
Induction is a Data Mining technology in which a
collection of objects are scanned to identify
similarities. These similarities are reported in a rule
form, i.e. If condition then conclusion. Rule
induction uses two important parameters in identifying
rules; the support and confidence. |
| |
|
|
| Support and
Confidence |
|
The
"Support" is the number of rows that support a
given rule. The "Confidence" is the percentage
of data rows where the rule is true, to all relevant
rows. |
| |
|
|
| Example |
|
For example,
if there are 50 rows where Product is Jacket, and 40 of
which have High Profit. We can infer the following rule:
If Product =
Jacket then Profit = High
This rule is supported by 50
rows and has a confidence of 80% (40/50)
|
| |
|
|
| Facts |
|
Facts are
another representation of IF THEN Rules that is more
suitable for data mining. The above example will be
written as:
Most Sales
where Product = Jacket have Profit = High
|
| |
|
|
| Exceptions |
|
Exceptions
are cases where a few rows share a common property that
is different from the majority. These exceptions may
represent interesting cases or may represent errors in
data. An example is:
Only 1% of
Sales where Product = Jacket have Profit = Very High
|
| |
|
|
| Numerical Values |
|
There are
more opportunities for finding patterns in a table if it
contains columns that have repeated values. For example
columns like Year, Month, State, Season, Class, Color,
etc. Numerical data where there are no repeated values
must be classified into levels like High, Medium and Low.
SuperQuery offers a number of
features to automate this process using its Wizards,
Range Columns, and Summary Tables.
|
| |
|
|
| More Information |
|
For more
information refer to our data mining White Paper. |
Office Edition
