Subscriber Data Mining in Telecommunication
3 likes1,860 views
This document outlines a project to perform subscriber data mining on telecommunication data for business reporting and decision making. The project aims to perform customer segmentation, new campaign planning, customer relationship management, and churn prediction. It involves collecting call detail records, preprocessing the data, designing a data mart and OLAP cube, applying clustering and other data mining algorithms, and visualizing the results. The goals are to help telecommunications companies better understand customer behaviors and improve customer retention through targeted campaigns.
Subscriber Data Mining in Telecommunication
- 1. Supervisor Babu Ram Dawadi Co-Supervisor Manoj Ghimire Subscriber Data Mining for Business Reporting and Decision Making in Telecommunication Project Member Bishal Timilsina Bishnu Bhattarai Narayan Kandel Niroj Karki 1
- 2. PRESENTATION OUTLINE 1. Introduction 2. Block Diagram 3. Data Collection 4. Data Preprocessing & Loading 5. Data Mart Design 6. OLAP Design 7. Data Mining Algorithm 8. Visualization 9. Result & Conclusion 10. Reference 2
- 3. INTRODUCTION Problem Statements Product based strategies rather than customer based strategies Problem on CRM Ignorant about customer behaviours Objectives Customer Segmentation New Campaign Customer Relationship Management Reporting To know more about customer, their call detail Churn Prediction 3
- 5. DATA COLLECTION 1. csv format 2. Txt format 5
- 6. ETL PROCESS Extract Extract tab separated data from txt file using bash shell & regular expression Transform Male to m Female to f Business_call to 1 Non_business_call to 0 Load Load data to mysql database using python script 6
- 8. OLAP DESIGN Purpose: Slice, Dice, Roll up, Drill Down operation Design Basis: 4 dimension representation 8
- 9. DATA MINING 1. RFM Methods R Recency (x axis) F Frequency (y axis) M Monetary (z axis) Step Attribute Selection & K-means Clustering 9
- 10. DATA MINING … 2. Two Phase Clustering 10 Objective Customer Segmentation How? 1st clustering -> Diamond, Gold, Silver … 2nd cluster -> Demographic cluster Now compare cluster based on attribute value
- 11. DATA MINING … 2. Two Phase Clustering 11 Objective Customer Segmentation How? 1st clustering -> Diamond, Gold, Silver … 2nd cluster -> Demographic cluster Now compare cluster based on attribute value
- 12. DATA MINING … 2. Two Phase Clustering 12
- 13. DATA MINING … 3. Gaussian Distribution 13 Objective Churn Prediction through call diameters How? Predict customer with value outside 90% confident range Accuracy? With increase in data size-> accuracy increase
- 16. VISUALIZATION 3. Time Series Visualization 16
- 18. VISUALIZATION 5. OLAP Visualization … 18
- 20. RESULT & CONCLUSION With the implementation of this software, telecommunication will be able to know more about customer & their call behavior Customer Segmentation help them 1.To maintain effective customer relationship management 2.To launch specific offers focusing on specific groups Alert them about customer churn behavior 20
- 21. LIMITATION & FURTHER ENHANCEMENT Limitation Data load time is high Our System Isn’t customizable for all query Further Enhancement Customer Segmentation accuracy could be improve by including customer life time value & apriori algorithm Reporting tool could be made more general & flexible Competitor Analysis 21