Telecom Churn Case Study

Problem Statement

With 21 predictor variables, the task is to predict whether a particular customer will switch to another telecom provider or not, known as churning or not churning, respectively.

About the Data

Variable Name	Meaning
CustomerID	The unique ID of each customer
Gender	The gender of a person
SeniorCitizen	Whether a customer can be classified as a senior citizen
Partner	If a customer is married/in a live-in relationship
Dependents	If a customer has dependents (children/retired parents)
Tenure	The time for which a customer has been using the service
PhoneService	Whether a customer has a landline phone service along with internet service
MultipleLines	Whether a customer has multiple lines of internet connectivity
InternetService	The type of internet services chosen by the customer
OnlineSecurity	Specifies if a customer has online security
OnlineBackup	Specifies if a customer has online backup
DeviceProtection	Specifies if a customer has opted for device protection
TechSupport	Whether a customer has opted for tech support or not
StreamingTV	Whether a customer has an option of TV streaming
StreamingMovies	Whether a customer has an option of Movie streaming
Contract	The type of contract a customer has chosen
PaperlessBilling	Whether a customer has opted for paperless billing
PaymentMethod	Specifies the method by which bills are paid
MonthlyCharges	Specifies the money paid by a customer each month
TotalCharges	The total money paid by the customer to the company
Churn	Target variable indicating if a customer has churned or not

Data Analysis and Modeling

Extensive exploratory data analysis (EDA), data scaling, and modeling were performed.

Models Used:

• Logistic Regression
• Decision Tree
• Random Forest

Results:

• Logistic Regression: Achieved 80% accuracy.
• Decision Tree and Random Forest helped identify tenure as the most significant predictor influencing churn.

Libraries Used

• warnings
• pandas
• sklearn
• matplotlib
• seaborn
• statsmodel