customer-churn's Introduction

Customer Churn Prediction

Predict if a customer will churn or not using Machine learning

In this repository, we have performed the end to end Exploratory Data Analysis, and idenfitied the characteristics of the customers that are more likely to churn, and I have used them wisely to create a model, and lately, have deployed the model using flask and heroku.

Approach :

Performed some data cleaning and feature engineering on raw data.
Selected 6 best features to deploy.
fitted multiple classification model and finally selected the stacked classification model.
Saved the model in a .pkl file and.
Later used the same model in the flask app and for frontend used HTML, CSS and Bootstrap.
Deployed the whole project on ``Herokuand usedGoogle Analytics` for tracking users.

How to run?

To run the app you need to download this repository along with the required libraries. and you have to the app.py file.

after running app.py open http://127.0.0.1:5000

Document Structure

Personal Finance 
│
|---- Data
|   |-- ML_models
|   |   |--
|   |
|   |-- preprocessed_data.csv
|   |-- WA_Fn-UseC_-Telco-Customer-Churn.csv
|
|---- images
|   |-- Churn-Prediction_Trim.gif
|
|---- notebooks 
|   |-- models
|   |   |-- analyseModel.py
|   |   |-- hyperparameterTuning.py
|   |   |-- *.ipynb
|   |-- *.ipynb
|
|---- static 
|   |-- images
|   |   |-- favicon
|   |   |   |-- *.png
|
|   |-- styles
|   |   |-- layout.css
|   
|---- templates
|   |   |-- index.html
|   |   |-- layout.html
|   |   |-- prediction.html
|
|---- .gitignore
|---- app.py
|---- LICENSE
|---- Procfile
|---- README.md
|---- requirements.txt
|---- runtime.txt

Technologies used :

python library - numpy, pandas, seaborn, matplotlib, flask, plotly, sklearn, pickle, xgboost
version control - git
backend - flask
concept - Machine Learning