Classification-Project-Report

---

Primary Goals:

• Determine key drivers of churn from Telco Dataset.
• Determine the most viable ML model and accurately predict churn with it.

---

Procedure

• Acquire telco dataset from MySQLWorkbench
• prepare data
  • remove excessive columns
  • fill null values
• explore date in search of key drivers for churn
• try to answer leading questions
• how often does churn occur?
• is churn rate dependent on monthly charge?
• does having benefits reduce churn rate?
• Develop a model to accurately predict if customers will churn
• • Use identified drivers to predict churn
  • evaluate models using training and validate sets
  • Determine optimal model on accuracy
  • Run test set on optimal model

---

Data Dictionary

Feature	Definition
Senior Citizen	If a customer is a senior citizen, 0 = No, 1 = Yes
Tenure	The amount of months a customer has been with or is currently with company
Monthly Charges	Amount a customer is charged monthly
Total Charges	Cumulative amount a customer has paid
Gender	If a customer is male or female, 0 = Female, 1 = Male
Has Partner	If a customer has a partner, 0 = No, 1 = Yes
Has Dependents	If a customer has dependents, 0 = No, 1 = Yes
Has Multiple Lines	If a customer has multiple lines, 0 = No, 1 = Yes
Contract	Type of contract customer has, 0 = Month-to-month, 1 = One year, 2 = Two year
Internet Service	Type of Internet Service customer has, 0 = No internet service, 1 = DSL, 2 = Fiber optic
Has Automatic Payment	If a customer has automatic payment, 0 = No, 1 = Yes
Has Amenities	If a customer has a majority of amenities from (tech_support, online_security, paperless_billing, streaming_movies, online_backup, streaming_tv, device_protection), 0 = No, 1 = Yes
Has Internet Service	If a customer has internet service, 0 = No, 1 = Yes
Churn (Target)	If a customer has churned, False = No, True = Yes

---

Steps to Replicate

• Clone this repo.
• Acquire the Telco data from MySQLWorkbench
• Put the data in the file containing the cloned repo.
• Run notebook.

---

Conclusions

• About 26% of customers churn currently
• Out of that 26%, we can significantly reduce churn rate by moving customers to a yearly contract instead of a monthly one.
• Having internet service actually made churn rate worse.
• Stronger drivers may become more prevalent if we can isolate yearly contracts from monthly ones.