Customer Churn Prediction

 Using 

Machine Learning

Problem Description:

• Customer churn is a term used when a customer decides to stop using the services of the business. Businesses do customer churn analysis all the time because it is very helpful for a company if they learn which customers are about to leave.

• This aim of this project is to train a machine learning model on the available data to train a machine learning model that will predict with a high accuracy which customers are about to churn, which in turn will help the business owner in making useful marketing decisions.

• You can find the original project details on Kaggle: Dataset.

• Data Fields are as shown below:

• Data fields:

• 1) State, string - 2 letter code of the US state of customer residence. 

• 2) Account Length, numerical - Number of months the customer has been with the current telco provider.

•  3) Area_code, string="area_code_AAA" where AAA = 3 digit area code.

•  4) International_plan, (yes/no) - The customer has international plan. 

• 5) Voice_mail_plan, (yes/no) - The customer has voice mail plan. 

• 6) Number_vmail_messages, numerical - Number of voice-mail messages. 

• 7) Total_day_minutes, numerical - Total minutes of day calls. 

• 8) Total_day_calls, numerical - Total number of day calls. 

• 9) Total_day_charge, numerical - Total charge of day calls. 

• 10) Total_eve_minutes, numerical - Total minutes of evening calls. 

• 11) Total_eve_calls, numerical - Total number of evening calls. 

• 12) Total_eve_charge, numerical - Total charge of evening calls. 

• 13) Total_night_minutes, numerical - Total minutes of night calls. 

• 14) Total_night_calls, numerical - Total number of night calls. 

• 15) Total_night_charge, numerical - Total charge of night calls. 

• 16) Total_intl_minutes, numerical - Total minutes of international calls. 

• 17) Total_intl_calls, numerical - Total number of international calls. 

• 18) Total_intl_charge, numerical - Total charge of international calls 

• 19) Number_customer_service_calls - numerical. Number of calls to customer service 

• 20) Churn, (yes/no). Customer churn - target variable.

• 
  

in the exploratory data analysis part some of the main findings were :

firstly - target variables were imbalanced so this needed to be taken care of while building the models 

Higher number of samples are churning when the day_charge is higher then 40.

Most of the churned_samples can be seen when the number of voice mail messages is higher than 15

as the customers may get annoyed if they were sent a lot of mails 

then we found out that some features were closely correlated so we  removed it in order to eliminate redundancy