Ce diaporama a bien été signalé.
Nous utilisons votre profil LinkedIn et vos données d’activité pour vous proposer des publicités personnalisées et pertinentes. Vous pouvez changer vos préférences de publicités à tout moment.

Data mining in telecommunications industry

An overview of some of the applications of data mining in the field of telecommunications industry

  • Identifiez-vous pour voir les commentaires

Data mining in telecommunications industry

  1. 1. Data Mining in Telecommunications Industry Prepared by: Alhassan Hammoud, Issa Memari, Soha Yazji
  2. 2. Outline  Introduction  Types of Telecommunication Data  Call Detail Data  Customer Data  Data Mining Applications  Marketing/Customer Profiling  Customer Segmentation  Customer Churn Prediction
  3. 3. Introduction  Telecommunications industry generates a tremendous amount of data  Details of every call  Details of every customer  Billing details  Services  Mine the data for profitable knowledge  Difficulties with mining telecommunications data  Scale  Rarity  Raw data
  4. 4. Types of Telecommunications Data: Call Detail Data  Descriptive information about every call is saved as a call detail record  MTN generates about 110 call detail records daily for every 100 customers  Call detail records include sufficient information to describe the important characteristics of each call  Originating phone number  Terminating phone number  Date and time of the call  Duration of the call  Not directly used for data mining  Extract knowledge at customer level
  5. 5. Types of Telecommunications Data: Call Detail Data  Summarize call detail records associated with a single customer  Summary variables (over some time period P)  Average call duration  Percentage of no-answer calls  Percentage of calls to/from a different area code  Percentage of weekday calls (Sunday - Thursday)  Percentage of daytime calls (9am – 5pm)  Average number of calls received per day  Average number of calls originated per day  Number of unique area codes called during P
  6. 6. Types of Telecommunications Data: Customer Detail Data  Telecommunications companies maintain a database of information on their customers  Name  Address  Service plan  Credit  Billing and payment history  Customer data is often used in conjunction with other data to improve results
  7. 7. Data Mining Applications: Marketing/Customer Profiling  Information mined from customer detail and call detail data can be used for marketing purposes  Syriatel’s SHABABLINK offer  Reduced calling fees for calls to people in one’s calling circle  Add entire circles of customers  Establishing and marketing international calling plans  Privacy concerns
  8. 8. Data Mining Applications: Customer Segmentation  Customer segmentation is often approached with cluster analysis  K-means clustering is commonly applied to customer profile data Class Characteristics 1 High values for international call minutes and data usage 2 High values for SMS and data usage 3 High values for all call variables 4 High values for all variables 5 Average values for all variables
  9. 9. Data Mining Applications: Customer Churn Prediction  Customer churn involves a customer leaving one telecommunication company for another  Significant problem for telecommunications companies  Example: companies offering incentives for signing up  Mine historical data to predict customer churn  Take action
  10. 10. Data Mining Applications: Customer Churn Prediction  Binary classification problem  Commonly used data mining techniques  Naïve Bayes classifiers  Multilayer perceptron classifiers  Decision tree classifiers  Evaluation criteria  Highly imbalanced dataset  F-measure
  11. 11. Data Mining Applications: Customer Churn Prediction  Confusion matrix  Percentage of positive cases caught: recall = 60/100  Percentage of correct positive predictions: precision = 60/200  F-measure: 𝐹 = 2 × 𝑃𝑟𝑒𝑐𝑖𝑠𝑖𝑜𝑛×𝑅𝑒𝑐𝑎𝑙𝑙 𝑃𝑟𝑒𝑐𝑖𝑠𝑖𝑜𝑛+𝑅𝑒𝑐𝑎𝑙𝑙 Predicted Negative Predicted Positive Negative Cases TN: 9760 FP: 140 Positive Cases FN: 40 TP: 60
  12. 12. References  AlOmari D., Hassan M.M. (2016) Predicting Telecommunication Customer Churn Using Data Mining Techniques. In: Li W. et al. (eds) Internet and Distributed Computing Systems. IDCS 2016. Lecture Notes in Computer Science, vol 9864. Springer, Cham.  Weiss, G.M. (2005) "Data mining in telecommunications" Data Mining and Knowledge Discovery Handbook. Springer US.

×