Telco Churn Classification Project

Description

This project aims to answer why the customers churn for the telco company. This will be done through a jupyter notebook report that will explore the different features that are affecting churn. We will also be creating predictive model that tells both the likelyhood that a customer will churn and whether or not the customer will churn. We will also be making a presentation that walks through how the model works. Finally, we will have python files that contain functions that will deliver future data to our model in the same form that we have trained the model to receive.

Goals

• Notebook with report of the findings
• .csv file with predictions
• .py files:
  • acquire.py
  • prepare.py
• Presentation
• Completed README

How to Reproduce

First clone this repo

acquire.py

• Must include env.py file in directory.

  • Contact Codeup to request access to the MySQL Server that the data is stored on.
  • env.py should include the following variables
    • user - should be your username
    • password - your password
    • host - the host address for the MySQL Server

• As long as you have the env file then get_telco_data() will do the rest on it's own.

prep.py

• prep_telco_data() will split, clean, and encode the data for the preparation stage.

• prep_all_data() will clean and encode a dataframe for using in modeling

telco_churn_report.ipynb

• This contains the report of what is causing churn and a walkthrough of the pipeline to creating a model to predict whether a customer will churn.

Planning

• Acquire
  • acquire.py
    • Get data from MySQL: telco_churn: customers
    • Determine which tables join on
    • Make function to acquire data into a pd.DataFrame
  • Notebook
    • Demonstrate acquire.py
    • Summarize data
    • Plot Distributions
• Prepare
  • prepare.py
    • Split data
    • Handle Missing Values
      • Handle datatype issues
    • Encode strings
    • Scale data
    • Add new feature tenure_years
    • Create feature that combines phone_service and multiple_lines
    • Create feature that combines dependents and partner
    • Look into merging other variables
      • streaming_tv and streaming_movies
      • online_security and online_backup
  • Notebook
    • Explore missing values and document takeaways/action plans for handling them.
    • Document takeaways
    • Explore datatypes
      • Adapt types or data values as needed to have numeric representation of each attribute.
    • Demonstrate prepare.py works by running it
• Exploration
  • Answer key questions
• Feature Engineering
• Modeling
  • Notebook

Data Dictionary

Columns	Definition
customer_id	customer number
gender	gender identity
senior_citizen	0 = senior citizen, 1 = not senior citizen
partner	undefined
dependents	undefined
tenure	total months as customer
phone_service	yes, no = phone service
multiple_lines	yes, no, not phone service
internet_service_type_id	no, yes, no internet serivice
online_security	yes = subscribed, no = not subcribed, or no internet service
online_backup	yes = subscribed, no = not subcribed, or no internet service
device_protection	yes = subscribed, no = not subcribed, or no internet service
tech_support	yes = subscribed, no = not subcribed, or no internet service
streaming_tv	yes = subscribed, no = not subcribed, or no internet service
streaming_movies	yes = subscribed, no = not subcribed, or no internet service
contract_type_id	refers to contract type
paperless_billing	no = sent by mail, yes = electronic notification
payment_type_id	id of method of payment
monthly_charges	current monthly charge
total_charges	total charges since becoming a customer
churn	yes = no longer a customer, no = still a customer
contract_type	Month-to-month, Two year. One year
internet_service_type	None, DSL, Fiber optic
payment_type	method of payment