The Fifth Elephant 2019

Gathering of 1000+ practitioners from the data ecosystem


Story of Building a Telecom Data Analytics Solution

Submitted by sawinder kaur (@sawinder) on Monday, 15 April 2019

Preview video


Telecom data is quite complex - consisting of hundreds of continuous and categorical variables that capture the details of millions of users consisting of plans, services, roaming, phone/SMS usage, revenue, and, cost, etc. Through interactions with customer leadership, we arrived on the business objective of our solution as optimizing the existing plans and services and maximizing the profit. We used statistics and ML-based feature selection techniques and combined with domain knowledge to recommend the features to the Subject Matter Experts (SMEs). Using our solution the SMEs were able to identify the end-user plans that were not profitable and able to come up with strategies to maximize the profits, tailor existing services and offer new services. In this crisp talk, I will share the details on how data science played a role in assisting the telecom customer to derive valuable insights.


In this talk, I will share my knowledge and experience in building a data science solution for telecom domain.
Here is an outline of my talk:
1. Define business objectives and translating the customer requirement into a data science problem for telecom domain.
2. Explain the data - Telecom end-user data contains thousands of attributes. What are the attributes required for the business goals? How do we use domain knowledge and data science to do feature selection?
3. What’s the data processing pipeline? How to keep telecom SMEs involved in feature selection and identifying the opportunities to maximize the profits?
4. Details of categorical feature selection techniques and the ML model that assisted the telecom SMEs.
5. The takeaway message from Telecom data analytics.

Speaker bio

PhD in Applied Mathematics with multiple years of work experience in data science and ML. Currently working as a Data Scientist at SAP Labs. Prior to SAP, worked at Shell India ML group as a data scientist.


Preview video


  • Anwesha Sarkar (@anweshaalt) Reviewer 7 months ago

    Thank you for your submission. Submit your preview video and slides by 21st March(latest). It helps us to provide a fair evaluation to the proposal and close the review process.

  • sawinder kaur (@sawinder) Proposer 7 months ago

    21st April?

    • Anwesha Sarkar (@anweshaalt) Reviewer 7 months ago

      Yes :) 21st April 2019 it is.

  • Zainab Bawa (@zainabbawa) Reviewer 7 months ago

    Thanks for the submission Sawinder. Who is the target audience for this talk? What will participants learn from this talk? What will participants, who don’t work with telecom data or are not concerned with telecom data, get out of this talk?

  • sawinder kaur (@sawinder) Proposer 7 months ago

    @Zainab - Thanks for your comments.
    - Target audience - beginners and analysts in data science who have not build a data science solution/product in the industry
    - Using telecom data as an example but the talk will not be completely focus on the domain, my goal is to abstract the learnings and present to the audience. Specific learnings for the non-telecom audience is (1) How to perform feature selection and feature reduction for high-dimensional datasets specifially for categorical variables? (2) What design considersations should be used while building a data science solution for a B2B use case
    - This proposal is for a Crisp Talk - 20 mins session.

    • Zainab Bawa (@zainabbawa) Reviewer 7 months ago

      The slides shared don’t reflect this takeaway.

Login with Twitter or Google to leave a comment