Result:
A social network model is saved in SAP HANA for future applications and a table with social
network analysis results for each of our existing customers. The following step adds this social
network data to the previous data manipulation.
Manipulating the Data for Customer Churn Analysis a Social Network
This section will outline the steps needed to add the social network data to our existing data manipulation. The
final predictive model is built on this combined data set.
After saving your social data set to SAP HANA, return to the home screen of SAP Predictive Analytics
Select the
Data Manager
, then
Load an Existing Data Manipulation
Connect to your SAP HANA instance, logon as a <Domain User>
Select the previous data manipulation: DM_CHURN_CUSTOMERS
In the
Data Manipulation Editor
, select the
Merge
tab
ADD a new merge using CUSTOMER_ID from the source table and SENT from the social data target
table SAP_RDS_PA_TELCO.SII_CUSTOMERS_SOCIAL
Choose
Next
Save the new data manipulation as DM_CHURN_CUSTOMERS_SOCIAL
Result:
The data manipulation DM_CHURN_CUSTOMERS_SOCIAL is saved to SAP HANA and
contains the added data from the social network analysis. In the next section, the final predictive
model is built with this data manipulation
Building the Classification/Regression Model for Customer Churn Analysis
From the home screen of SAP Predictive Analytics, select the
Modeler
section
Select
Create a Classification/Regression Model
In the
Select a Data Source
window, choose
Browse
to select your SAP HANA instance using the
logon <Domain User>
Select the
Browse
button again to select the data set that will be used for predictive modeling, your
previously saved data manipulation DM_CHURN_CUSTOMERS_SOCIAL
Note that data manipulations will appear at the top of the list of options
If your data manipulation is built with the prompt as specified above that prompt asks for the last date
of the training period
After specifying the last training date, choosing
Next
begins modeling
Choose
Analyze
to read in a description for your data set,
Choose
Next
after confirming that the data description is accurate
The data description was created previously in the data manipulation steps. You should not
need to change the description
In the
Selecting Variables
screen, set
CHURNED_IN_M1
as your target variable and exclude the
variables
CUSTOMER_ID, ZIP_CODE, CHURNED_IN_M2, CHURNED_IN_M3, KCOMINDEX
and
NUMBER
from your analysis,
Choose
Next
If desired, unselect
Enable Auto-Selection
Choose
Generate
to build your customer churn model with SAP Predictive Analytics
After reviewing model results, SAVE this model as SII_CHURN_MODEL_FINAL to the SAP HANA
schema SAP_RDS_PA_TELCO
Result:
The final churn model is ready to be deployed using current customers data to identify
potential churners.
Applying the Model (Optional Exercise)
After saving the model, select
Apply Model
from the
Run
section
