Skip to content

Latest commit

 

History

History
137 lines (96 loc) · 6.6 KB

README.md

File metadata and controls

137 lines (96 loc) · 6.6 KB
Raw

Codeup Project


Project Telco


Learn to Discern What Turns Customers to Churn

  • Discover drivers of churn of Telco customers

  • Use drivers to develop a machine learning models to identify drivers of churn

  • Churn as a customer ending their contract or not renewing their contract with Telco

Project Description

Telco, a telecommunications enterprise, provides a wide array of services catering to a diverse clientele. This endeavor delves into the exploration of distinct elements influencing customer churn. The goal is to ascertain whether any of these factors amplify or diminish the probability of customers discontinuing their services.

Project Goal

  • Identify the drivers of churn among Telco customers.

  • Develop a machine learning model to classify customer churn, distinguishing between contract renewals and contract terminations.

  • Enhance our understanding of which customer attributes contribute to or mitigate customer churn.

Initial Thoughts

I hypothesize that the drivers of churn will likely involve dissatisfied customers. Specific services or the lack thereof might be influencing customers to churn.

The Plan

  1. Data Acquisition

    • Obtain data from the Codeup MySQL Database.

  2. Data Preparation

    • Create new engineered columns from the existing data.

  3. Data Exploration

    • Explore the data to identify potential drivers of churn by answering initial questions:
      • Is churn independent from payment type?
      • Is churn independent from internet service type?
      • Is churn independent from paperless billing?
      • Are there variations in churn based on monthly charges?

  4. Model Development

    • Utilize the insights gained from the exploration to build predictive models.
    • Evaluate model performance on training and validation data.
    • Select the best-performing model based on accuracy.
    • Validate the chosen model using the test data.

  5. Conclusion Drawing

    • Summarize the findings and insights.

Data Dictionary (Exploration)

Feature Values Definition
index: customer_id Alpha-numeric Unique ID for each customer
mailed_check_payment True=1/False=0 Whether customer is/has feature name
e_check_payment True=1/False=0 Whether customer is/has feature name
credit_card_payment True=1/False=0 Whether customer is/has feature name
bank_transfer_payment True=1/False=0 Whether customer is/has feature name
two_year_contact True=1/False=0 Whether customer is/has feature name
one_year_contract True=1/False=0 Whether customer is/has feature name
internet_service_type True=1/False=0 Whether customer is/has feature name
month_to_month_contract True=1/False=0 Whether customer is/has feature name
automatic_payments True=1/False=0 Whether customer is/has feature name
churn (target) True=1/False=0 Whether customer is/has feature name
dsl_internet True=1/False=0 Whether customer is/has feature name
paperless_billing True=1/False=0 Whether customer is/has feature name
fiber_optic_internet True=1/False=0 Whether customer is/has feature name
streaming_tv True=1/False=0 Whether customer is/has feature name
tech_support True=1/False=0 Whether customer is/has feature name
device_protection True=1/False=0 Whether customer is/has feature name
online_backup True=1/False=0 Whether customer is/has feature name
online_security True=1/False=0 Whether customer is/has feature name
multiple_lines True=1/False=0 Whether customer is/has feature name
phone_service True=1/False=0 Whether customer is/has feature name
kids True=1/False=0 Whether customer is/has feature name
married True=1/False=0 Whether customer is/has feature name
senior_citizen True=1/False=0 Whether customer is/has feature name
streaming_movies True=1/False=0 Whether customer is/has feature name
tenure_years Numeric Normalized (0 - 1) Tenure normalized with MinMaxScaler()
total_add_ons Numeric Normalized (0 - 1) Add ons normalized with MinMaxScaler()
tenure_months Numeric Normalized (0 - 1) Tenure normalized with MinMaxScaler()
monthly_charges Numeric Normalized (0 - 1) Charges normalized with MinMaxScaler()
total_charges Numeric Normalized (0 - 1) Charges normalized with MinMaxScaler()

Steps to Reproduce

  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.
    • RandomState 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.

Conclusions

Takeaways and Key Findings

  • Customers without tech support are churning more than those without
  • Payment type, especially electronic check, is a significant driver of churn.
  • The influence of fiber optic internet service on churn is surprising given its high speed.
  • Paperless billing increases churn, with many churn cases having it enabled.
  • Churn rates tend to rise with higher monthly charges.

Recommendations

  • Investigate and address issues related to electronic check payments.
  • Analyze potential problems with fiber optic internet service.
  • Offer tech support for free or cheap if it means retaining customers.

Next Steps

  • Given more time, delve into the reasons behind the high monthly charges contributing to customer churn.
  • Tune the hyperparameters to potentially find a better tuned model