nicesoul / babushka-churn

БабУшка [Babushka] - an application to predict a customer churn rate, a studying project at https://becode.org

БабУшка /Babushka/ churn prediction

An application to predict a customer churn rate - a studying project at BeCode.

The Mission

One financial institution is interested in analyzing its client database to increase the revenue generated from credit cardholders. They are concerned about some customers closing their bank accounts after accepting products from other institutions.

The churn rate is above 15% and increasing, so the CEO urges the marketing team to start a marketing campaign for clients retention.

My mission is:

• Predict clients who have a propensity to close their bank accounts with the financial institution.
• Find possible groups of clients and define their characteristics. That will help the marketing team to design custom-made campaigns to increase customer retention.
• Build a dashboard with data insights and KPIs.

What is Churn Rate?

The churn rate, also known as the rate of attrition or customer churn, is the rate at which customers stop doing business with an entity. It is most commonly expressed as the percentage of service subscribers who discontinue their subscriptions within a given time.

More info

Mission objectives

• Build machine learning models for classification.
• Build machine learning models for clustering.
• Select the right performance metrics for a model.
• Identify the optimal number of clusters and be able to describe them.
• Tuning parameters of the model for better performance.
• Describe the results from unsupervised learning.
• Online deployment of a dashboard with data insights.

Instuctions

Currently, this repo consists of 3 folders:

• model = all the machine learning models used for predictions.
• dashboard = files for building a dashboard for the app.
• docker = files needed to containerize my app and deploy it.
• docs = files related to a website with documentation.

Requirements

Before running please be sure to install the required libraries.

pip install -r requirements.txt

Note that some libraries might require Microsoft C++ Build Tools preinstalled.

Tested to work for Python 3.9.9 x64 (Windows 10)

Contact

In case of any question, please, refer to documentation (currently under development)

Here is my contact page

If you want to check out other variations of this application, built by my groupmates, please, let me know - I will redirect you to the best ones.