28
PREPAID CHURN MODEL With Oracle Data Mining Necdet Deniz Halıcıoğlu [email protected] September 21, 2010

P REPAID C HURN M ODEL With Oracle Data Mining Necdet Deniz Halıcıoğlu [email protected] September 21, 2010

Embed Size (px)

Citation preview

PREPAID CHURN MODELWith Oracle Data Mining

Necdet Deniz Halıcıoğ[email protected]

September 21, 2010

Agenda

Conclusion

SVM Model

Existing Mining System in Turkcell

Churn Prediction

About Turkcell Technology

Data Mining with ODM

Agenda

Conclusion

SVM Model

Existing Mining System in Turkcell

Churn Prediction

About Turkcell Technology

Data Mining with ODM

Turkcell Technology has more than 15 years of development experience with its solutions applied and proven at leading operators in more than 10 countries.

2009

More than 10 years of experience in Turkcell ICT

TTECH Center was put into serviceHC: 255 engineersFocus: Turkcell Group

Focus: Turkcell & Telia Sonera Group + Regional SalesHC: 360 engineers

TTECH was formed with 44 engineers in TÜBİTAK-MAM Technological Free ZoneFocus: Turkcell

Focus: Turkcell & Telia Sonera GroupHC: 321 engineers

2008 Today20071994 - 2006

About Turkcell Technology

Areas of Competency

From assisting the operation of network resources to improving business oriented intelligence, TTECH’s experts provide an expanding portfolio of packaged and custom solutions for telecom network operators.

Network Services & Enablers

SIM Asset & Services Management

Mobile Marketing

Mobile Internet & Multimedia

Business Intelligence & Support Systems

Turkcell Technology IMS Group

More than 10 years of BI experience in Telecommunications industry

Designed, Built and Running one of the largest data warehouses in telecom industry

Team of more than 100 highly talented professionals and consultants

Has a proven record of success in BI operations Flawless operation, providing data for finance and even for NYSE

Early adopter of the new BI trendsComplex Event Processing, Text Mining, etc.

Agenda

Conclusion

SVM Model

Existing Mining System in Turkcell

Churn Prediction

About Turkcell Technology

Data Mining with ODM

What Makes Churn Prediction So Crutial?

Everybody faces the same difficulties…

Competition

Forming Customer Loyalty

High cost of customer acquisition

Optimizing budget for customer retention

People don’t want to hear any more

Basics of Churn Prediction

Churn prediction starts with turning an abundance of data into valuable information and continues as a cyclic process.

Data

Preparation Preprocessing Mining

Information

Action

Define variable pool

Perform mining ETL

Attribute Importance

Normalization Outlier

Detection Missing Value

Cleanup

Build Test Apply

Success Criteria

• • Customer Loss

• Useless Action• Customer

Annoyance•

0/0

0/11/1

1/0

Agenda

Conclusion

SVM Model

Existing Mining System in Turkcell

Churn Prediction

About Turkcell Technology

Data Mining with ODM

Too much manual effort: A new project for every new mining activity

SAS licensing

Not leading, but lagging the business

Administrative overhead of distributed mining environment Network overhead Decoupled process monitoring Data quality problems

Pain Points About Existing Mining System

E-DWH DM-DWH SAS Server End Users

Approach in Existing Churn Model

Attribute Selection with Human Expertise

Perform ETL

Build many models in serial with different• Algorithms• Hyperparameters

Choose best model manually

Replace the existing model with the best model for churn

prediction

Agenda

Conclusion

SVM Model

Existing Mining System in Turkcell

Churn Prediction

About Turkcell Technology

Data Mining with ODM

Motivations

Building an automated mining framework based on our Oracle

database experience instead of maintaining manual mining model

cycle.

No extra licensing cost (under ULA).

High speed (close to real time) mining with database embedded

mining.

Centralized mining activity monitoring & administration.

Give a Try to Oracle Data Mining

Oracle

Our Proposal for Data Mining Framework

Feed all customer attributes possible

Let AI to filter important ones

Train Oracle SVM models with

selected attributes

Externalize those models for APPLY

Choosing Attributes with Attribute Importance

--Perform EXPLAIN operationBEGIN

DBMS_PREDICTIVE_ANALYTICS.EXPLAIN(data_table_name => 'census_dataset',explain_column_name => 'class',result_table_name => 'census_explain_result');

END;/

--View resultsSELECT * FROM census_explain_result;

COLUMN_NAME EXPLANATORY_VALUE RANK-------------- ----------------- ---- IN_REF_NUMDAYSSINCELASTREFILL .141200904 1DT_SUB_ACTIVATIONDATE .028200303 2IN_MNP_PORTINTENURE .026178093 3NM_SUB_ACTIVATIONREASON .025882544 4IN_MNP_TCELL_TENURE .025279836 5...

Our Top 5 After AI

Top

5 by

AI

Number of days since last refill

Activation Date

Port in Tenure

Subscriber Activation Reason

Subscription Period in Turkcell

--Perform PREDICT operationDECLARE

v_accuracy NUMBER(10,9);BEGIN

DBMS_PREDICTIVE_ANALYTICS.PREDICT(accuracy => v_accuracy,data_table_name => 'census_dataset',case_id_column_name => 'person_id',target_column_name => 'class',result_table_name => 'census_predict_result');

DBMS_OUTPUT.PUT_LINE('Accuracy = ' || v_accuracy); END;/

--View first 10 predictionsSELECT * FROM census_predict_result WHERE rownum < 10;

PERSON_ID PREDICTION PROBABILITY ---------- ---------- ----------- 2 1 .418787003 7 0 .922977991 8 0 .99869723 9 0 .999999605 10 0 .9999009

5 rows selected.

Build & Apply the SVM Model

No need to perform manual attribute processing in many cases

EDP : Embedded data preparation

ADP : Automatic data preparation

PL/SQL or Java based code generation

SAS to ORACLE model import• Eliminates data Movement• Eliminates data duplication• Preserves security

Other Remarks on ODM

Agenda

Conclusion

SVM Model

Existing Mining System in Turkcell

Churn Prediction

About Turkcell Technology

Data Mining with ODM

Creating the Case Table

Variable Pool(400 variables)

JOIN MONTH(N)=MONTH(N+1)

Filtered Variable Pool

PREPAID and INDIVIDUAL and

(ACTIVE or MOC-BARRED)

Historic Churn Table

CASE TABLE

Building the SVM Model

CASE TABLE• 400 Attributes

• Unique Identifier• Target Churn Value

ATTRIBUTE IMPORTANCE

FEB DATA MAR CHURN

CASE TABLE(180 ATTRIBUTES)

MAR DATA APR CHURN APR DATA MAY CHURN MAY DATA JUN CHURN

COMBINE DIFFERENT DATASETS

BUILD SVM MODEL

ODM on Oracle Exadata v2

o Initially we have used a large Solaris (100+ UltraSparc 7 cores and 640 GB memory) box to build our first SVM models:• It took 29 hours to complete

model build & apply.o On Exadata this reduces to a few

hours.oMainly due to enormous

improvement in data preparation stage.

Agenda

Conclusion

SVM Model

Existing Mining System in Turkcell

Churn Prediction

About Turkcell Technology

Data Mining with ODM

Churn prediction over various customer groups is and will be the focus of Turkcell

Embedded data mining with ODM is Faster More Robust (due to stability of SVM algorithm) Easier to automate Easier to manage

To Sum Up

Thanks for his contribution

Data & Information Technologies

Hüsnü Şensoy, VLDB [email protected]

To learn more on SVM theory

Turkcell Technology Research and DevelopmentTÜBİTAK MAM Teknoloji Serbest BölgesiGebze – KocaeliTURKEY

' : +90 (262) 677 40 007 : +90 (262) 677 40 018 : www.turkcelltech.com

THANK YOU!