Databases Reference
In-Depth Information
the features of VisMiner previously described in the tutorial. If, however,
the objective is to apply the data mining knowledge discovered to future
decision-making activities,
then your best previously constructed models
should be employed.
Suppose that Wheels has decided to decided to purchase additional potential
customer data from Tell All and to use the ensemble model to choose
individuals from the new dataset to receive their proposed promotional offer.
How should it proceed?
The dataset PotentialBuyers.csv contains a list of 5,000 new individuals from
Tell All.
Open PotentialBuyers.csv.
View the summary statistics.
The new dataset contains all attributes found in the original data except for
the Buyer column. We will not know if individuals from this list are going to buy
until we send them the offer and they have a chance to try out the car. Who
should receive this offer?
Drag PotentialBuyers.csv over the previously created ensemble model and
release.
Select “Generate predictions”.
A new dataset named “Predicted:PotentialBuyers.csv” is created.
To avoid such a cumbersome name, rename this dataset “PredBuyers”.
View the summary statistics for PredBuyers.
The newly generated dataset has all of the columns of the original plus three
more: BuyerPredicted, NProb, and YProb.
BuyerPredicted is the class (Y or N) predicted by the model for each
observation. Notice that there are 4,000 predicted as “N” and 1,000 as “Y”.
NProb and YProb are the likelihoods that the observation is “N” and “Y”
respectively.
Use either the parallel plot or the Control Center's filtered dataset
option to create a subset named “LikelyBuyers” of individuals where
YProb
0.90.
>
Save the LikelyBuyers dataset.
View the summary statistics.
