Aggregation
Celatro offers several techniques whereby various pattern matching algorithms can be aggregated together. One mechanism for this involves writing code: Celatro's robust Object Oriented Design (OOD) makes it easy for sofware developers to quickly create customized pattern matching approaches for specific data mining needs.
More significantly, aggregation of Celatro components can be accomplished at run-time. Celatro's component-based architecture lets you create, on the fly, composites that have exotic new behaviors different from their constituents.
The following demo features one type of aggregation found in Celatro: run-time specification of how measure scores are combined to produce an aggregate score. Here, a user can enter an expression that indicates how he or she wants to combine various algorithm scores to produce a final result.
Instructions: Enter an expression using mathematical operators to combine results. Specify which data field you want using brackets and the field-name or zero-based index.
|
Example Expressions |
|
1.5 * [0] + 3 * ["Country"] |
Multiply the measure of the first match by 1.5, the second by 3, and add up the results. |
|
3 * ["Song"] ^ 2 |
Square the value for "Song" and multiply the result by 3. |
|
IIF( ["Song"] > 0.5, ["Song"], 0) |
Use the value for "Song" if it is greater than 0.5, otherwise score as zero. |
|
Subject Matter:
|
|
|