In the financial industry, the prediction of the default is a main concern for all banks around the world. Machine Learning is a way to improve the process of decision making when someone asks for a loan.
This work aimed to create a machine learning engineering project to predict the default. Continuos Integration (CI), Continuos Deployment (CD), and Continuos Training are applied in this work.
Throughout this work, the code is able to retrain machine learning models and deploy them automatically. This approach overcomes data and model drifts.
The following architecture showcases how to create end-end a machine learning project with CI-CD-CT.
GitHub is mainly used for the CI.
GitHub Actions is mainly used for the CT-CD.
Azure is used as hosting service of the API through Container Registry and App services. Meanwhile, the data, metrics, and models are read and saved in Blob Storage.
The final idea is to automate all tasks in the machine learning modeling. Data drift and model drift over the time are going to degradate the performance of a single model. The final idea is to avoid this phenomenon with a strong architecture.
The CI allows to add new pieces of code quicly.
The CD allows to new changes in the code easily.
The CT allows to train periodically new machine learning models based on how data change over the time.
Public code is available in the following GitHub repo.Pujilí, Cotopaxi, Ecuador
sebitas.alejo@hotmail.com
© Sebastián Sarasti Zambonino. All Rights Reserved.
Designed by HTML Codex
Edited by Sebastián Sarasti and Angel Bastidas