MainaliB / Telco_Project

Telco_Project

Churn Predictors

Goals:

To create a classification model that can accurately identify the customers as someone that will churn or not using the TELCO data. To attain this goal, I will be identifying the drivers of the churn and create a model that will incorporate these drivers(variables) to accurately predict the customer’s likelihood to churn.

Data Dictionary:

• customer_id: unique identifier for each customer

• senior_citizen: (int) 1 indicates the customer is a senior citizen, 0 indicates they are not

• partner: (int) 1 indicates they have partners, 0 indicates they don’t have partners

• dependents: (int) 1 indicates they have dependents, 0 indicates they don’t have dependents

• tenure: (int) length customer has been with the company in months

• phone_service: (int) 1 indicates they have phone service, 0 indicates they don’t have phone service

• multiple_lines: (int) 1 indicates they have multiple lines, 0 indicates that they don’t

• online_security: (int) 1 indicates customers opted in for online security service, 0 indicates they havent

• device_protection: (int) 1 indicates customers have device protection, 0 indicates they dont

• tech_support: (int) 1 indicates customers have tech support, 0 indicates they dont

• streaming_movies: (int) 1 indicates customers have streaming movie service, 0 indicates they dont

• paperless_billing: (int) 1 indicates customers have enrolled in paperless billing, 0 indicates they havent

• monthly_charges: (int) monthly charge of a customer

• total_charges: (float) total charges customers have paid

• churn: (int) 1 to represent customers that have churned, 0 represent they havent

• partner_dependents: (bool) to represent if customers have both partner and dependents

• phone_and_multiple_lines: (bool) to represent if customers have both phone and multiple lines

• streaming_tv_movie: (bool) to represent if customers have both streaming tv and movie

• online_security_and_backup: (bool) to represent if customers have both online security and back up

• payment_auto: (int) 1 indicates customers have automatic payment, 0 indicates they have not

• month-to-month: (int) 1 indicates customers are with month to month, 0 indicates they have other form of contract

• one year: (int) 1 indicates customers are with one year, 0 indicates they have other form of contract

• two year: (int) 1 indicates customers are with two year, 0 indicates they have other form of contract

• DSL: (int) 1 indicates customers have DSL, 0 indicates they have other form of internet

• fiber optic: (int) 1 indicates customers have fiber optic, 0 indicates they have other form of internet

Project Planning

Data Acqusition:

• Acquire data from CodeUp database. You are required to have credentials to access the data base.
• Create a cache in your local repository in order to prevent from having to access the database every time you re-work on the project
• Create acquire.py for reproducibility.

Data Preparation:

• Summarize data
• Check for duplicates, removed duplicates if present
• Check for null values, missing values, and get count of them
• Check the data types
• Perform feature engineering
• Perform data types conversion/Use Dummy variables
• Drop redundant/useless columns
• Split data into Train, Test, and Validate
• Create prepare.py

Data Exploration:

• Explore various features and relationships
• Vizualize data
• Impute missing values
• Test different hypothesis

Model Building

• Try various algorithms(Logit, Random Forest, KNN, Decision Tess) on Train data
• Perform model evaluation
• Pick top 3 perfomring models
• Validate top 3 models on validate data and evaluate their performance
• Pick best performing model on validate data
• Use the top model with test data set and evaluate your model

Conclusion

• Summarize the process
• Summarize findings
• Make recommendations
• Road Ahead?