Telco Churn Calculator

• 

  Data Source - Input Variables Definitions

• 

  Dashboard

• 

  Result for specific User Inputs

Inspiration

Predicting the churn rate of new subscribers is crucial to the success of new and growing businesses, where customer retention takes precedence. Hence, an easy and simple approach to predict had to be explored.

What it does

The product simply predicts whether a customer churns or not, given the input parameters.

How we built it

The project is divided into. multiple models Random Forest Classifier and Neural Network, which can be switched when needed to produce the binary output.

Accomplishments that we're proud of

Getting it done in a short amount of time.

What we learned

Docker is great tool to have a containerized environment with all the required AI/ML packages and libraries made available without the hassle of dealing with dependencies and OS requirements. It helped us focus on the development of our project rather that compromising valuable time in satisfying packages and system requirements.

Challenges

One area that we wanted to explore was the integration of Tensorboard into the docker container that would allow custom configuration of a prebuilt model. Something that would resemble a UI with model parameters, allowing for the model to be re-trained and predict using the new model weights.

Built With

• fastapi
• html
• keras
• numpy
• pandas
• python
• requests
• scikit-learn
• tensorflow
• uvicorn