Telco-Churn-Prediction

This is a churn classification project on a fictitious company Telco prepared by IBM. Dataset link : https://www.kaggle.com/datasets/blastchar/telco-customer-churn/data

In this repository I have tried to build a well structured ML project, maintaining the directory hirarchy as well as model tracking for different experiments. At the end the model is deployed in Streamlit. You can try it out here: Live demo link

A quick sneak peek to the app :

How to reproduce the results ?

• step 1 : open a terminal in your local project directory.
• step 2 : run git clone https://github.com/Dion11235/Telco-Churn-Prediction/
• step 3 : run conda create -p churnenv python=3.9 to create the environment
• step 4 : run cd Telco-Churn-Prediction to move into the cloned directory.
• step 5 : run pip install -r requirements.txt to install all the dependencies.
• step 6 : run python src/pipeline/train_pipeline.py to run the experiments. Dont forget to delete the folder mlflow_experiments_log before running this, otherwise there will be duplicate experiments.
• step 7 : run mlflow ui --backend-store-uri=mlflow_experiments_log to check out the awesome mlflow dashboard!

Results :

Achieved F1 score 0.865 and AUC : 0.86, ROC curves for different experiments look like this -

If you like the work, give a star 🌟. Thanks in advance! 😄