Black Lives Matter. Please consider donating to Black Girls Code today.

Telco Customer Churn Analysis and Prediction

Hello everyone, I developed a Dash application! Here:

Heroku: https://lnkd.in/dxQYSTA
GitHub: https://lnkd.in/dPpEeUQ

This application allows you to predict Telco Customer Churn and analyze the dataset which is published by IBM.

I look forward to your feedback. You can view all codes and modelling notebook on GitHub.

Thank you :slight_smile:

3 Likes

Really nice!!

What are some inputs that cause the model to predict retention / “not churn”? I entered some parameters that I thought would be relatively “safe” but it still predicted churn!

1 Like

Thank you so much :slight_smile: @chriddyp ,

XGBoost model can estimate feature importance, so we can interpret dominant features for prediction. Unfortunately SVM (if not linear) is black box.

For example: keep all input values and just increase TotalCharges to 240 and Tenure to 40.

Here is the XGBoost feature importance: