| |
| The UHP-DMSM
approach addresses the problems above with a proprietary customer-centric
datamart topology, and three proprietary software tools (VISTAR™,
MiningSolve™, and DELTA™). |
| |
| The Ultra-High Performance Data Mining
ProtocolSM has the following benefits: |
| |
- Proven world-class lifts: 200 to 300% over conventional approaches
(case study available upon request).
|
- Speed: Up to millions of data mining models built, scanned and
validated per week. A "model" is a true, separate algorithm
by transform, by subalgorithm, by iteration, by criteria combination.
|
- Unparalleled distributed computational capacity: The UHP-DMSM
can connect computers over a LAN, WAN, or the Internet to create
a virtual supercomputer capable of solving problems far faster
than achievable by traditional servers or supercomputers.
|
- Validation using hold-back sampling: The only truly valid way
to test model performance is to apply it to "hold-back"
respondents, respondents who were not used in developing the model,
but for whom we already know their "disposition." (This
is termed hold-back sample validation.)
|
- Learns from previous successes by storing them in a rules repository.
|
- Recommends the best models: Compares lifts at different campaign
depths.
|
