Data Mining and CRM in Telecommunication Industry

MT
by murat temur0 views·Updated Apr 21, 2014
Powered by AI

Need to present on this topic?

Create your own polished presentation in minutes. Just describe what you need.

Free to try, no credit card required

Transcript

SYE 492 ENGINEERING PROJECT

"BIG DATA"

The new competitive advantage!

DATA MINING

BIG DATA AND CRM

What We Did?

  • Anticipating and preventing customer churn:

Prepaid customer churn for youth segment

  • Targeting the right customer with the right offer:

Our new online billing service

  • Understanding influencers in future selection:

Forecast (Genç Turkcell) university students’ pre-paid and post-paid mobile card selection probabilities

The core CRM issues include

  • Customer acquisition
  • Customer retention
  • Cross-sell/Up Sell
  • Maximizing Lifetime Customer Value

Data Mining and CRM in Telecommunication

Industry

INTRODUCING COMPANY

IMPORTANT BUSINESS PROBLEM : THE CHURN

Churn Model For Prepaid Customers

  • Turkcell is an innovative telecom operator providing communication solutions to telecom operators, corporations and households in the areas of data, voice and video.
  • To ease and enrich the lives of customers with communication and technology solutions.
  • To grow in core mobile communication business through increased use of voice and data

Churn Model For Prepaid Customers

CONTENTS

ODMtelcosetup.sql script

Churn represents the loss of an existing customer to a competitor.

A prevalent problem in retail:

  • Mobile phone services
  • Home mortgage refinance
  • Internet services

Model to predict churn and score the model to identify high-valued customers who are at-risk for churn

-Number of minutes per customer, broken out by tariff-month combination

-Number of minutes per customer, rolled up to the tariff level

M1,M2...

to tweak settings

and rerun for common scenario

Attributes which have the strongest contribution towards churn

-Length of service

-Age

-The number of weekend usage minutes

nvl (mins,avg(mins)over(partitionby cust_id, tariff))

Missing value replacement

Customer identifier and the associated probability of churn

CUSTOMERS

SERVICES

CDR_T

REVENUES

CHURN_RISK

rolls up the total revenue per customer

  • Introducing Company
  • Scope of Project
  • Big Data
  • Data Mining and CRM
  • Applications
  • Conclusion

Forecasting Online Billing Service Customers

CONCLUSION

Today, data mining is used to analyze mass databases in telecommunication industry for having meaningful output

CRM can be successfully practiced with effective, efficient and reliable data mining methods

Forecasting University Students’ Prepaid and Postpaid Mobile Card Selection Probabilities

1st target: 16-25 years old

  • high smartphone penetration
  • early adopters of new technologies

Barriers for adoption:

  • dependency on parents

SCOPE OF PROJECT

Solution:

  • possibility to limit spending (smartphone function)

Affecting Factors

  • Demographic factor : students’ age, gender and civil status, class
  • Social factor : number of family members and number of students studying in a family
  • Economic factor : income, communication expenses, total expenditure and family income

n: Mass size

N: The main mass

P: Postpaid usage probability

Q: Prepaid usage probability (1-P)

Z: Test value in %(1-α) confidence

α :Confidence level d :Tolarance

P=Q=1/2 we take 1000 sample and use result of 774

  • Churn Model For Prepaid Customers

Independent Variables

MPE: Mobil communication expenses

NOS: Number of students studying in a family

G: Gender

C: Class

To achieve this -> Logit model, one of the binary choice model

26.171 questionnaires applied to the university students by Genç Turkcell Youth Club

:

Logit model cdf:

Postpaid: 1 , Prepaid : 0

Girl : 1 , Boy : 0

P:probability of occurence

Y: two dependent variable

e=2.711828.

  • Forecasting Online Billing Service Customers
  • Forcasting University Students’ Prepaid and Postpaid Mobile Card Preference

Thanks for Attention

Team Members

Ahmet Can RAMAZANOĞLU

Murat TEMUR

Supervisor

Asst. Prof. Dr. Dilek KAPTANOĞLU

Jury Members

Asst. Prof. Dr. Dilek TÜZÜN AKSU

Prof. Dr. Linet ÖZDAMAR

https://www.wevideo.com/hub/#media/ci/150574973

12.02.2014

1

Girl student post-paid preference probability:

MPE=17,00 TL, NOS=2 and C=3 with solving equality:

=>

Post-paid customers :boys ratio is higher, NOS is small, students is at higher classes and have higher mobile phone expenses.

2

Automatically sifting through large amounts of data to find previously hidden patterns, discover valuable new insights and make predictions

3

10,000 Customers

  • CUSTOMERS
  • SERVICES
  • CDR_T
  • REVENUES
  • CHURN_RISK