Telco_Report

Project Description:

The telecommunications enterprise Telco faces a significant challenge with customer retention. This project aims to unearth the underlying factors contributing to customer churn, employing advanced machine learning techniques to identify and predict patterns of customer attrition. The ultimate goal is to enable Telco to implement preemptive strategies to enhance customer loyalty and reduce churn, with the findings and predictive model being presented to the lead data scientist for strategic decision-making.

Goals:

• Discover drivers of churn.
• Construct a machine learning classification model that can accurately predict churn.
• Present findings and process to lead data scientist, as well as recommendations.

Initial Hypotheses:

• Churn is related to charges and potentially some services.

Questions pertaining to the data:

• What is the relationship between monthly charges and churn?
• Does internet (or lack of) affect churn?
• Do any of the internet services have a particular impact on churn?
• What contract types cause higher churn?

Data Dictionary:

Variable	Description
index	Customer_ID, the identification code associated with each customer
gender	Male or Female, the gender of each customer
senior_citizen	Yes or No, whether or not the customer is a senior citizen
married	Yes or No, whether or not the customer has a spouse
children	Yes or No, whether or not the customer has children
tenure_months	Integer, the amount of months the customer has been with Telco
paperless_billing	Yes or No, whether a customer has signed up for paperless billing
monthly_charges	Float, each customer's monthly charge
total_charges	Float, each customer's total charge to current date
churn (target)	Yes or No, whether a customer has ended his or her contract with Telco
contract_type	The type of contract the customer has with Telco (Month-to-Month, 1 Year, or 2 Year)
internet_service_type	The type of internet the customer has with Telco (Fiber, DSL, or none)
payment_type	How a customer pays for his or her service
streaming	Details on which streaming services a customer may be subscribed to
phone_lines	Details on a customer's phone service
protection	Details on which online protection services a customer may have
support	Details on which online support services a customer may have
Additional features	Encoded values for categorical data for the sake of modeling

Project Pipeline:

• Acquisition
  • Acquire data from local CSV or Codeup's SQL database
• Preparation
  • Rename columns for clarification
    • married
    • children
    • tenure_months
  • Create engineered columns from existing data
    • streaming
    • phone_lines
    • protection
    • support
• Exploration
  • Explore the data to identify potential drivers of churn
    • What is the relationship between monthly charges and churn?
    • Does internet (or lack of) affect churn?
    • Do any of the internet services have a particular impact on churn?
    • What contract types cause higher churn?
• Modeling
  • Encode the data
  • Use feature and hyperparameter selection to build predictive classification models
  • Test predictive models on train and validate
  • Isolate best model to run on test dataset
• Delivery
  • Draw conclusions
  • Present findings

Reproduction of Findings:

1. Clone this repository
2. If you have access to the Codeup MySQL DB:
   • Save env.py in the repository with user, password, and host variables.
   • Ensure the env.py has the appropriate database connection.
   • random_state of 123 is predefined in the functions
   • Run the notebook.
3. If you don't have access:
   • Request access from Codeup.
   • Follow step 2 after obtaining access.

Notes: With the exception of acquire.py and env.py, all the .py files can be rebuilt from the contents of their respective notebooks.

Key Findings

• Monthly charges play a significant role in whether a customer will churn or not.
• There is a greater proportion of churn from customers with internet as opposed to those without internet.
• Monthly customers churn most from fiber optic internet.

Recommendations

• Offer more deals to bring monthly charges down for month-to-month contract types
• Explore any potential issues with the fiber optic service that may result in greater churn.