belladu0201 / ids706_beibei-du-project4

IDS706_Beibei-Du-Project4

• Create a Microservice that returns a JSON payload and performs a Data Engineering related task
• Push tested source code to Github and perform Continuous Integration with Github Actions (or similar SaaS Build service)
• Configure Build Server to Deploy Changes on build (Continuous Delivery)
• Create realistic API (reference here: Data Engineering: Chapter 5 aws chapter for pragmatic ai.)

Overview

In this project, I am using the FastAPI to deploy some data enginerring Tasks. I take a dataset from Kaggle about the fetal health and I want to discover what are the features that will classify/predict the fetal_health status and using machine learning to achieve a high accuracy on testing set. I considered the following models: KNN, Random Forest, Logistic regression, XGBoost. In the output that I received, XGBoost performed the best with an accuracy score of 94.13%. The confusion matrix is shown below:

In the next step of this project, I try to randomly select 10 rows from the dataset and find the mode fetal_health status. By doing that, I have write out scripts in cli.py and main.py. It can be run through the command in the following. I run the command several times and the most frequent health status for the fetal is 1.0, which is the Normal Status.

Implementation

• Install the packages in the requirements.txt and set up for FastAPI
• Code for the project and functions to perform the tasks
• Perform Continous Intergration on Github Actions (main.yml)
• Deploy Continous Delivery

How to Use

• Use the command uvicorn main:app --reload to run the server

Reference

https://www.kaggle.com/datasets/andrewmvd/fetal-health-classification https://github.com/noahgift/functions-from-zero2