ML Mini-Project

---

Telecom Retention | Customer Churn Analysis | Churn Prediction

---

One of the most common issues, widely faced by any kind of industry, is retaining their customers, which is a challenging task (be it the telecom industry, gaming industry, etc).

With the aim of customer retention, this project helps in identifying and understand the customer behaviour within the Telecom Industry, in order to identify the churning customers.

Understanding the Different Types of Churn:

The concept of Customer Churn is not just limited to the shift made by a customer from availing the services/products of one firm to another. There are several ways in which a company may suffer a loss due to customer churn. There is:

• Service Churn (for eg. downgrading from a monthly to a weekly subscription)
• Tariff Plan Churn (for eg. downgrading from a Rs. 100 plan to a Rs.50 plan)
• Product Churn (for eg. changing from postpaid to prepaid)
• Usage Churn (for eg. inactive or zero usage )
  

These are specific to the telecom industry.

Different Churning Segments:

As per the decision cycle of a typical subscriber, four churn segments come into the picture:

1. Conditionally Loyal: Customers who have thought about churning > Are not locked in by their contract > Still decided to stay
2. Conditional Churner: Customers who have thought about churning > Are not locked in by their contract > Decide to leave because they found a better option
3. Lifestyle Migrator: Customers who have thought about churning > Are not locked in by their contract > Decide to leave because their needs changed
4. Unsatisfied Churner: Customers who have thought about churning > Are not locked in by their contract > Decide to leave because they're not satisfied

Overview of a Data-Science Approach to manage churns:

Capturing and Analysing Data > Reporting and Prediction > Engaging and Taking Action

The ultimate goal is to identify the customers who are likely to be churned, along with their category of churn, so that we are able to tackle their issues and retain them as customers.

For the EDA part:

Refer the ML_ChurnPrediction_EDA.ipynb file

For the Full EDA + Model Building:

Refer the ML_ChurnPrediction(1).ipynb file

Original Dataset:

WA_Fn-UseC_-Telco-Customer-Churn.csv

Cleaned Dataset After Data Preprocessing (Used for Model Building):

tel_churn.csv

Model File:

model.sav

App Code File:

app.py

Deployment WebPage Format:

index.html

Model Deployed At:

https://amritaveshin.github.io/ML_TelecomChurnPrediction/