CTS FWR5-3105SFP-CW-AC-DR Series User manual

User Guide for Predictive Analytics Cloud Image - Trial Edition
SAP Store –Trial Edition

Table of Contents
1INTRODUCTION......................................................................................................................................3
2PREREQUISITES.....................................................................................................................................3
3SYSTEM ACCESS...................................................................................................................................3
4PREDICITIVE SCENARIOS (FRONT-END COMPONENTS).................................................................5
4.1 Banking....................................................................................................................................................5
Customer Attrition Analysis.......................................................................................................................5
4.2 Consumer Products .............................................................................................................................10
Brand Sentiment and Sales Analysis .....................................................................................................10
Demand Data Analysis...........................................................................................................................12
Product Fulfillment and Optimization......................................................................................................14
4.3 Finance ..................................................................................................................................................18
Company Performance Analysis ............................................................................................................18
Late-Payment Management ...................................................................................................................18
Customer Cash Collection Analysis .......................................................................................................19
4.4 Manufacturing.......................................................................................................................................22
Customer Demand and Inventory Management.....................................................................................22
Overall Equipment Effectiveness............................................................................................................23
Asset Breakdown Analysis .....................................................................................................................24
Maintenance Cost Analysis ....................................................................................................................28
4.5 Portfolio & Project Management.........................................................................................................34
Project Profitability Analysis....................................................................................................................35
4.6 Retail......................................................................................................................................................42
Market Basket Opportunities ..................................................................................................................42
Customer Loyalty Programs...................................................................................................................43
Store Clustering......................................................................................................................................45
4.7 Sales & Marketing.................................................................................................................................47
Customer Segmentation.........................................................................................................................48
Market Segmentation..............................................................................................................................49
Market Campaign Success.....................................................................................................................50
Product Recommendation ......................................................................................................................51
Pipeline and Revenue Forecasting.........................................................................................................53
4.8 Telco.......................................................................................................................................................56
Churn Modeling and Offer Recommendation.........................................................................................59
Post-Paid Analysis..................................................................................................................................64
Rotational Churn Detection.....................................................................................................................65
Multi-SIM Detection ................................................................................................................................68
5HANA MODELS (OPTIONAL BACK-END COMPONENTS)................................................................73
6OPEN SOURCE R INCLUDING INSTALLATION (OPTIONAL)...........................................................82
7SECURITY ASPECTS............................................................................................................................82
8APPENDIX..............................................................................................................................................82
8.1 Further Documentation........................................................................................................................82
8.2 SQL Scripts ...........................................................................................................................................82

1 INTRODUCTION
Users learn how to use SAP Predictive Analytics tools in the context of SAP HANA with a few pre-built
scenarios in this trial edition for the cloud. Sample data is available so that the users can understand how
the tools could be utilized. Predominantly the emphasis is on SAP Predictive Analytics tool (both Expert
Analytics and Automated Analytics roles), SAP Lumira and SAP UI5 (HTML 5). SAP HANA is showcased
for housing the data and modeling the data.
2 PREREQUISITES
The trial edition is available in the SAP Store. Users access the trial landscape from their laptop through a
remote desktop connection.
3 SYSTEM ACCESS
Process Steps
In the SAP Cloud Appliance Library, choose Instances to display the list of available instances. Then
choose the Connect operation for your instance and follow the instructions. Alternatively you can
manually configure a connection in the following way.
Choose the name of the running Error! Reference source not found. instance. The system opens
the Instance dialog box with the properties of the solution instance. The IP Addresses area displays
the access details of the solution instance.
Copy the IP address of the running instance.
Use the SAP Logon New functionality and enter the details of the new system, then start an SAP GUI
connection.
Parameter
ID
Parameter
Value
Note
IP address of
the instance
x.x.x.x
To be retrieved from SAP Cloud Appliance Library when viewing
the details of the solution instance.
SID
00
System ID of Predictive Analytics Trial Edition on SAP HANA
Remote Desktop
Group
IP
User
PW
All
IP address of your
personal instance
Administrator
The master password is used for
accessing the system. It is provided by the
user during the creation of the solution
instance in SAP Cloud Appliance Library.
HANA Studio: <your system>, Instance <your instance>
Group
Backend SAP
HANAServer
Name
User
PW
All
VHCALHDBDB
SYSTEM
master password
Client Tools (SAP Predictive Analytics –Expert Analytics role and Automated Analytics role)
Tool
Group
Backend SAP
HANAServer
Name
User
PW

Predictive Analytics
All
VHCALHDBDB
RDSPAUSER
master
password
Predictive Analytics
(for Sales&Markeing
use cases)
All
VHCALHDBDB
RDS_CRM_DS
USER
master
password
Connect to Remote Desktop
Start Remote Desktop by following the menu Start > All Programs > Accessories > Remote Desktop
Connection
Now the Remote Desktop is launched.
Enter the following information:
Computer: IP addresses of your personal instance
User: Administrator
Choose Connect.
Enter the following information:
Password: Master Password
Choose OK.
Connect to Data Sources (ODBC)
Start Data sources (ODBC) by following the menu Start > All Programs > Administrative Tools > Data
Sources (ODBC)
Click on System DSN tab and click on the Add button
In the new data source pop-up dialog, select HDBODBC as the data source and click Finish

Now enter the data source name, data source description as vhcalhdbdb and Server:Port as
vhcalhdbdb:30015
Click connect
Enter user id and password as RDSPAUSER and master password
NOTE: Location of LUMs files on the instance
1. Once the instance is created and you have done a remote desktop, you get access to the files on the
remote desktop.
2. The Predictive Analytics LUMs files are stored in the D:\PA_Documents
3. Most of the LUMs files are already imported and you should find them when you launch “Expert
Analytics” from the SAP Predictive Analytics software. If you happen to not find the LUMs file that you
are looking to open, please import them from the location specified in step 2.
4 PREDICITIVE SCENARIOS (FRONT-END COMPONENTS)
4.1 Banking
For the Banking industry, as part of the solution, we have a pre-built predictive model for one use case, the
customer attrition analysis. In this use case, we focus on analyzing the customer attrition trends using 2 different
approaches, identifying the customers like to leave
Using the Automated Analytics tool in Predictive Analytics
Using the Expert Analytics tool in Predictive Analytics
We have built a generic data structure and seeded sample data so that compelling predictive models could be
built. Most of the fields in the data structure are generic and which are used by many banking customers.
Customer Attrition Analysis
The following section describes configuration for the Customer Attrition Analysis scenario, which can be used
with either Automated Analytics or Expert Analytics. The generation and applying of this model will generate a
table in SAP HANA, which is then combined with Profile information and made visible via a database view in
SAP HANA. The data in this view can then be displayed using any tool that can read SAP HANA.
Note: Any SAP or non-SAP customer would be able to deploy or mimic the data structure, load the data and use
the pre-built models.
Automated Analytics (using the first approach)
Training the Model
Launch SAP Predictive Analytics
Choose the Modeler section
Select Create a Classification/Regression Model
Select Database from Select a Data Source
Select your SAP HANA instance
Logon to the <DOMAIN USER> account via the Browse button
Specify the Data Set that is used to Train your data model. The set is the table you populated with
your data: SAP_RDS_PA_BANK.CUST_ATTRITION

For more information on populating SAP_RDS_PA_BANK.CUST_ATTRITION, see the SAP
HANA Deployment for Banking on SAP Predictive Analytics Content Adoption rapid-deployment
solution (VD2) configuration guide that is part of this solution.
Choose Analyze
BUS_PARTNER is the only column with a Key value of 1. This value is generated by SAP
Predictive Analytics that identifies this column as the Primary Key of the table
On the next screen, enter Attrition as the Target Variable and make BUS_PARTNER the Excluded
Variable
The value of the key is meaningless in terms of the analysis
On the next screen, enter the Model name: ATTRITION_SAP_RDS_PA_BANK_MODEL
Choose Generate
A Model is created with the statistics and the metadata that are applied to the data set of customers to
predict whether or not they will attrite.
Review the Training the Model display to validate your Model.
The most important values are the quality indicator KI and the robustness indicator KR. For more
information about SAP Predictive Analytics values and indicators, see the Appendix.
On the Using the Model screen, verify that you are able to use the Model (see the following section on
the predict phase)
If you feel confident that this model is reflective of the additional data sets you are planning to apply
this model to for prediction purposes, Save this model for these future runs.
a. Expand the Save/Export band and then choose Save Model
b. Specify where you want the model saved (SAP HANA database, Text file, and so on) and
SAVE.
Once saved, models may be opened and reused. The folder location for the saved model is
selected when you open a saved model in SAP Predictive Analytics by selecting Load a Model.
If you feel less confident about your model because the Predictive Power (KI) and Predictive
Confidence (KR) on the screen are low (less than .9), then more modeling is required.
When additional modeling fails to indicate a trend or does not provide significant degree of
confidence in the results, review the data set used. If the data set is divergent or sparse, it
cannot support the generation of a robust model.
In Applying the Model, select your SAP HANA instance and logon to the <DOMAIN USER> account.
In the dialog box, select SAP_RDS_PA_BANK.VW_CUST_PROFILE
Result: The Model is ready for use as the basis for prediction. Continue to the next section for the
prediction phase.
Predicting
In SAP Predictive Analytics, the training and the prediction phases are a continuous process. To initiate the
Predict Phase, continue from the last step of the previous section.
From Run, select Apply Model
Specify the data set for predictions using your Model in the Application Data Set section.
a. Select Database
b. Browse and populate the dialog box with your SAP HANA instance
c. Logon to the <DOMAIN USER> account.
d. Choose Browse Data
e. In the dialog box, select SAP_RDS_PA_BANK.CUST_PROFILE (returning the data from the
table you previously loaded into SAP HANA)
If browsing the data doesn’t show this view, enter the name in the dialog box to use as a search
field

From Generation Options, select Decision for the Generation Options
Leave Apply as the Mode
Select where the generated output is stored by making the following entries in the Results Generated
by the Model section:
a. Select Database
b. Browse and populate the dialog box with your SAP HANA instance
c. Logon to the <DOMAIN USER> account.
d. Browse Data and in the resulting dialog box, select
SAP_RDS_PA_BANK.CUST_ATTRITION_RESULTS
The correct SAP_RDS_PA_BANK.CUST_ATTRITION_RESULTS table is all upper case
Select the Apply button to execute the Model
Choose the Delete option for the Update will attempt to merge the runs together dialog box, to remove
output rows from the previous run. When the cycle is complete, SAP Predictive Analytics returns
statistical information available under the Display bar.
Result: The predicted data is written back to the SAP HANA schema SAP_RDS_PA_BANK as a table
CUST_ATTRITION_RESULTS. To view the results in SAP Predictive Analytics, select the VW_
CUST_ATTRITION_RESULTS View, which joins the results with the original data.
Expert Analytics (using the second approach)
Training the Model
Select the LUMS for Banking_Customer_Attrition_Analysis - Train
a. Launch Expert Analytics in SAP Predictive Analytics. Open menu File and choose Import to
folder and import or if the LUMS file is already imported browse through the Documents list for
Banking_Customer_Attrition_Analysis - Train.lums file.
b. In the dialog box, enter the SAP HANA server details, your user name, and password
Verify the data
a. Choose Prepare panel to ensure that the data has loaded properly.
Switch to the Predict panel to view the predictive model
Adjust the HANA C4.5 - 4 independent cols Algorithm Component
a. Choose Configure Settings
b. Select the Properties/Output Information/Output Mode to Trend
c. Select Properties/Column Selection/Features for the independent columns (variables) for
analysis. By default, the following are enabled:
Average Tenure
Average Income
Average Savings Amt
Average Investment Amt
d. Choose Properties/Column Selection/Target Variable for the dependent column (variable) for
analysis. By default Attrition is enabled
e. Select Advanced/Tree Pruning/Minimum Split and enter the value 10.
f. Select Done
At this point, run the Algorithm Component (Run Till Here) before continuing the analysis. Choose Yes
to switch to the Results view for verifying the execution results
Adjust the HANA C4.5 - 3 independent cols Algorithm Component
a. Choose Configure Settings
b. Select Properties/Output Information/Output Mode to Trend

c. Select Properties/Column Selection/Features for the independent columns (variables) for
analysis. By default, the following are enabled:
Average Tenure
Average Income
Average Investment Amt
d. In the Properties/Column Selection/Target Variable, choose the dependent column (variable) for
analysis. By default, Attrition is enabled
e. Select Advanced/Tree Pruning/Minimum Split and enter the value 50.
f. Select Done
At this point, run the Algorithm Component first (Run Till Here) before continuing the analysis. Choose
Yes to switch to the Results view for verifying the execution results. The results should look similar to
HANA C4.5 - 4 independent cols Algorithm, though with less complexity in decision tree.
a. Select HANA C4.5 - 3 independent cols component. From Component Actions select Save as
Model.
b. Specify the Save Model Attrition_C4.5_3_indep_cols as Model Name
c. Select the Overwrite row and choose Save
Export Saved Model
a. From Components list, expand Models
b. Select Attrition_C4.5_3_indep_cols
c. From Component Actions, select Export Model
d. Choose Use this option to export data models to the SAP Predictive Analytics Archive (*.spar)
file
e. Change File Name to Attrition_C4.5_3_indep_cols.spar.
f. Select Save
Predicting
Once the model is trained, the PREDICT LUMS is used with SAP Predictive Analytics for predictive modeling
and the visualizations for the analysis using current data. The following steps must be executed to configure the
PREDICT LUMS.
Select the LUMS for Banking_Customer_Attrition_Analysis - Predict
a. Launch Expert Analytics in SAP Predictive Analytics. Open menu File and choose Import to
folder and import or if the LUMS file is already imported browse through the Documents list for
Banking_Customer_Attrition_Analysis - Predict.lums file.
b. In the dialog box, enter the SAP HANA server details, your user name, and password.
Verify the data
a. Choose Prepare panel to ensure that the data has loaded properly.
Switch to the Predict panel to view the predictive model. Banking_Customer_Attrition_Analysis -
Predict.lums include default saved model Attrition_C4.5_3_indep_cols imported into the analysis.
From Components list, select the ‘+’ button.
Select Import Model
a. Locate the saved Attrition_C4.5_3_indep_cols.spar, and choose Open
b. Select model Attrition_C4.5_3_indep_cols and choose Finish
Expand Models to select and drag Attrition_C4.5_3_indep_cols into Analysis screen
Adjust the HANA C4.5 - 3 independent cols Algorithm Component
a. Choose Configure Settings
b. Select in the Properties/Output Information/Output Mode to Trend

c. Select Properties/Column Selection/Features and choose the independent columns (variables)
for analysis.
d. Select: Average Tenure, Average Income, and Average Investment Amt identical to saved
model
Adjust the HANA Writer Data writer component
a. Select Configure Settings
b. Specify Schema Name, Table Type, and Table Name as appropriate for your site
Choose Run to execute the scenario
Choose Yes to switch to the Results view for verifying the execution results.

4.2 Consumer Products
For the Consumer products industry, we have pre-built scenarios for 3 use cases such as Brand sentiment and
sales analysis, Demand data analysis, Product fulfillment and Optimization. Depending on the use case and the
functionality that we are analyzing, we have picked up either the Automated Analytics or the Expert Analytics
approach.
Basically for the Demand data analysis and Product fulfillment and Optimization use cases, we have used the
leading SAP application Demand signal management as the data source and sample data sets are available for
the same. With regard to the Brand sentiment and sales analysis, we have built a generic data set and seeded
sample data set for building the predictive models.
Note: Any SAP or non-SAP customer would be able to deploy or mimic the data structure, load the data and use
the pre-built models.
Brand Sentiment and Sales Analysis
In this use case, we are trying to identify the brand sentiment during a major event and predicting the sales for
the upcoming similar events based on the brand value.
The following section describes configuration for the Brand Sentiment and Sales Analysis scenario. The
generation and applying of this model will generate a table in SAP HANA, which is then combined with Profile
information and made visible via a database view in SAP HANA. The data in this view can then be displayed
using any tool that can read SAP HANA.
Automated Analytics
Training the Model
Launch SAP Predictive Analytics
Choose the Modeler section
Select Create a Classification/Regression Model
Select Database on the Select a Data Source screen
Select your SAP HANA instance
Logon to the <Domain User> account via the Browse button
You now need to specify the Data Set to use to Train your data model. Choose
SAP_RDS_PA_CPG.SEM_MODEL_BASE
For more information on loading data into the SAP_RDS_PA_CPG.SEM_MODEL_BASE data
set, see Predictive Analytics for Consumer Products (K38) configuration guide that is part of this
solution.
On the next screen, choose Analyze
Assign Key as “1” against SALES_DATE, PRODUCT_NAME and EVENT
Check the Add Filter in data Set on the bottom left section of the screen.
Choose Next
Enter a filter for the date range of SALES_DATE using the format of YYYY-DD-MM, for example,
SALES_DATE <= 2014-02-15 and SALES_DATE >= 2012-02-13 (your range is based upon the date
range of your data set).
Enter a filter for the PRODUCT_NAME = PRODUCT 1/PRODUCT 2 (based on which product you
want to select for modeling)
On the next screen, use SALES_AMT as the target variable and
a. For training a model, which is based on event, use SMC1, SMC2, HLYDSSN, SALES_DATE
as Explanatory Variables Selected and exclude the rest of the variables