This project has been created to predict the churn of a customer and it's part of a challenge called Project-of-the-week that is being held by Datatalks.club.

• Churn-prediction app.
• Content
• Tools
• Dataset
• Enviroment
  • Create the enviroment
  • Activate the enviroment
  • Install dependencies
• Working on the notebook
  • Analysis conclusions
  • Jupyter Notebook into python code
• Flask Application
  • Create the app
  • Run the app
  • Send a request
• Front end
  • Install Streamlit
  • Working on design
  • Run the app
• Dockerfile
  • Build the image
  • Run the container
• Cloud
• After Course
  • Trying Gradio
  • Run the app

• Python
• Anaconda
• Pandas
• Scikit-learn
• Flask
• Streamlit
• Docker

Dataset's link

I'm going to use a conda enviroment for this project called churn-project.

conda create -n churn-project python=3.9

conda activate churn-project

pip install -r requirements.txt

Working on notebook: exploratory_analysis.ipynb Follow the notebook to see the results.

I tested three models:

• Logistic Regression
• Random Forest
• XGBoost

Random Forest and XGBoost have performed better than Logistic Regression and they have the same f1-score.

You can transform your notebook into python code by using the command:

jupyter nbconvert --to python exploratory_analysis.ipynb

Install Flask with the command:

pip install flask

Working on serve.py

python serve.py

I've created a simple script called serve_test.py that sends a request to the app and prints the response with this values:

REQUEST={
  "CreditScore": 597,
  "Geography": "Germany",
  "Gender": "Female",
  "Age": 35,
  "Tenure": 8,
  "Balance": 131101.04,
  "NumOfProducts": 1,
  "HasCrCard": 1,
  "IsActiveMember": 1,
  "EstimatedSalary": 192852.67,
  "Exited": 0
}

URL='http://localhost:9696/predict'

I'm going to use a Streamlit.io front end for this project.

pip install streamlit

Working on front in Front End

streamlit run front_end.py

FROM  python:3.9.12
LABEL Author, Esteban Encina

WORKDIR /app
COPY requirements.txt ./requirements.txt
## Install dependencies
RUN pip install -r requirements.txt
## Expose the port
EXPOSE 8501
COPY . /app
## Run the app
ENTRYPOINT [ "streamlit" , "run"]
CMD [ "app.py" ]

docker build -t churn-project .

docker run -p 8501:8501 churn-project

The application is deployed in Streamlit cloud.

Install gradio with the following command:

pip install gradio

I'll work with the file front_end_alternative.py

Run the app with the command:

python front_end_alternative.py