MiningSolve™
Feature Document
|
| ZZZ |
MiningSolve™
Release v.7.0 |
Unique
to MiningSolve™ |
| Types
of Analysis |
| Identify
best customer prospects based on cross-sell or any specified predictor |
ü |
ZZZ |
| Identify
profitable customers by current or lifetime value |
ü |
ZZZ |
| Predict
the current product life cycle |
ü |
ZZZ |
| Predict
customer loss or retention |
ü |
ZZZ |
| Enhance
the datamart so that every customer is scored with a) the key driver
importance and b) the needs based segment information so that future
database marketing models can include not only who should be called
for which product, but also what pitch to use |
ü |
ü |
| Algorithms
Supported |
| Use
SPSS™ via OLE for all model building |
ü |
ü |
| Supports
all relevant algorithms from SPSS™ |
ü |
ü |
| Genetic
algorithms*** |
ü |
ZZZ |
| Neural
nets*** |
ü |
ZZZ |
| Model
Set-Up |
| Select
multiple databases from any source to be used simultaneously for analysis*** |
ü |
ZZZ |
| Select
specific predictor and predicted variables to eliminate wasted resources |
ü |
ZZZ |
| Select
customers to be analyzed by customized region |
ü |
ZZZ |
| Select
customers to be analyzed by market segment |
ü |
ZZZ |
| Customize
the depth of analysis for algorithms, criteria, iteration, and transformation
to control analysis run hours |
ü |
ZZZ |
| Select
the methods and algorithms to use for data mining |
ü |
ü |
| Produce
reports of specific customers to target for products or services |
ü |
ZZZ |
| Models
can be customized by selected combination of algorithm, sub-algorithm,
criteria, iteration, or transform |
ü |
ü |
| Optimize
output by selecting specific optimization methods* |
ü |
ZZZ |
| Run-Time
Features |
| System
can review model performance in real time and re-resource to run more
of the more productive combinations of algorithms, sub-algorithms,
iterations, criteria or model transforms** |
ü |
ü |
| Manually
turn off or turn on certain combinations during the run |
ü |
ü |
| Fault-tolerance |
ü |
ü |
| Print
or save output to be used or modified later* |
ü |
ü |
| Speed
and Depth |
| Reengineered
OLE interface generates models over 1000 times faster than manual
models (produces over 5000 models per day using GLM models with 1000
cases, 20 variables on Pentium III 1.0 GHZ machine) |
ü |
ü |
| Ease
of Use |
| Does
not require advanced statistical knowledge |
ü |
ZZZ |
| Uses
"wizard" to set up a run |
ü |
ü |
| Technical
criteria entered by use of "slide controls" which define the depth
of the analysis |
ü |
ZZZ |
| Save
the settings for a specific analysis run to use repeatedly |
ü |
ZZZ |
| Expert
Systems Assistance |
| System
suggests appropriate models to run against the data* |
ü |
ü |
| Visualization |
| Correct
classification chart using hold back sampling, updated in real time |
ü |
ü |
| Chart
correct classification for all runs to compare run performance |
ü |
ü |
| Chart
run lift (correct classification over chance alone) for all runs |
ü |
ü |
| Status
of distributed datamining displayed for each connected computer |
ü |
ü |
| Display
of computers found on local area network |
ü |
ü |
| Software
platform |
| Visual
C++, 32 bit, MFC |
ü |
ZZZ |
| Stingray
Objective Grid for Tables* |
ü |
ZZZ |
| Architecture |
| VISTAR*
(VIrtual Subdatamart Testing And Reduction) creates thousands of subdatamarts
in order to find the reduced set of predictors that produces the highest
lifts |
ü |
ü |
| DELTA*
(Decision Enhancement of Lift Through Arbitration) builds models that
determine which of the best models from MiningSolve™ to "believe"
when these models disagree, resulting in higher lifts than any single
model alone |
ü |
ü |
| CCD
(Customer Centric Datamart) restructures the datamart so that every
record is a customer, not a household or an account |
ü |
ü |
| Distributed
Computing |
| Produce
multiple models simultaneously using networked computers |
ü |
ü |
| View
status of all connected servers from one client computer |
ü |
ü |
| Monitor
and execute runs from client computer |
ü |
ü |
| Take
advantage of unused cycles on idle networked computers |
ü |
ü |
| Remote
connect to computers over the Internet using an IP address |
ü |
ü |
| Consolidated
output files on client computer |
ü |
ZZZ |
| ZZZ
|
| *
= implemented by end of 4th quarter, 2003 |
| **
= completed by end of 1st quarter, 2004 |
| ***
= to be added by end of 2nd quarter, 2004 |
