Telco Churn
For this project, I discovered insights into customer habits that drive churn. Then, I created classification machine learning algorithms to predict if a customer would leave the company. I presented these findings with an emphasis on data storytelling.
To acquire the data, I connected to the Codeup SQL database using python and a query to join the tables together.
The data contained information on previous and current customers’ service type, monthly payments, and more. A breakdown of each service by internet, phone, and both is shown below. Not all options were available for each service, as most options were available for internet customers only.
The baseline model was the most common churn outcome - the customer does not churn. The final model selected was a Random Forest model, which was four percent more accurate than the baseline. A simplified visual of this model is shown below.
