|
What is
the Induction Engine?
Traditionally, testing multiple
data mining algorithms to find the best performing model required tremendous
domain expertise and months of testing. PredictionWorks has changed all
of that with its Induction Engine. Induction Engine automates the testing
of multiple data mining algorithms and chooses the appropriate sampling
size to complete in the allocated time. The power of data mining is harnessed
and automated within the Induction Engine.
What insight
will the Induction Engine give me?
Data mining allows users to predict outcomes of events in the near future.
This type of analysis differs from OLAP style reporting which focuses
on past events. Segmentation and clustering of customers is especially
useful to determine the intricacies of customer patterns. This information
gives insight into which customers will act on a campaign or personalized
offer. Predicting customer churn by scoring each customer with the probability
that they will leave, is the heart of any customer retention campaign.
How
do I use PredictionWorks Induction Engine?
PredictionWorks Induction Engine does all of the model generation based
on an uploaded Test and Train file. Most data modelers already have these
flat files that contain the customer or product profile in a denormalized
and pivoted format. Once the file is uploaded the user defines column
data types and which column contains the target (measurable result). The
difference between the Train and the Test file is that the Train file
contains the target column while the Test file is where the generated
model measures its accuracy.
Glossary: What is a
What are the
final Results?
PredictionWorks' Conclusion page contains more than just the predicted
results, but also more information to create a better model. Column ranking,
Decision Tree Visualization, and individual algorithm performance are
just a sampling of the information that the Induction Engine reports on.
This helps the user identify weaknesses within the model and the underlying
data infrastructure, which then allows for future improvements.
Here are sample test result files
on whether a Telecom company's customers will act on a particular Direct
Mail campaign offer. By performing this analysis this Telecom company
saved thousands of dollars on this one campaign by knowing which customers
they should send their message to. By doing this targeted messaging, they
were able to increase their response rate by 18 times.
Need
HELP
- Setting up a Train/Test file from your data?
- Interpreting your results?
- Integrating your models to your IT solution?
Contact us at 1-888-872-9566 or info@predictionworks.com
|
