Churn-Analysis---Telecom

The goal is to predict the propensity to churn of a customer from the service provider through the usage behavior using logistic regression.

Data Available

Voice Usage: Incoming / Outgoing Minutes of Usage for Local calls , STD calls, ISD calls, Local to other networks, STD to other networks, ISD to other networks.

Data Usage: 2G in MB 3G in MB

Payment data: Monthly Payment Data Total Bill, RentalCharge, Non Rental Charge

Subscriber information: Customer Type Account activation date Disconnection date (If applicable)

Customer care data Call to Customer Care Call type (Request / Query / Complaint / Feedback) Successive calls to customer care

Code Details:

File: SingleView.R Reads month wise input files, changes data types, calculates derived variables. Aggregates records for each ID and computes mean and variance of all variables. Generates churn_data / validation_data csv files.

File: churn_prediction.R Reads churn_data / validation_data Applies general logistic regression, plots AUC, computes confusion matrix.