Deploying ML Pricing Model on Flask Web App to AWS

Motivation

1. It's becoming more important to be equipped with Cloud technologies and be able to put machine learning models into production.

2. In the past, I was charged unexpected fees by AWS because I didn't configure it correctly at the time. A couple of my friends (one working for a company and the other at his own startup) had a similar issue and were charged a fair amount of AWS costs. I do not want to make a similar mistake at my future job (fingers crossed).

In this project, I wanted to deploy a web application that has a predictive model running on top of it by using AWS EC2 service (using the free tier this time!).

In other words, the main goal of this project was to:

Experience end-to-end data science project only in Python by deploying an ML serving web application to AWS EC2 instance

Here are the steps I have taken:

1. EDA, data cleansing, feature engineering, dimensionality reduction

2. Build a regression model and perform GridSearch for hyperparameter tuning and K-fold cross-validation

3. Build back-end by writing a python Flask server that uses the pickled model to serve http requests

4. Build a website front-end (html, css, javascript) that allows users to enter house information including home sqr ft area, # bed, # bath, and location.

5. Deploy the Flask web application to an AWS EC2 instance using Nginx.

---

Architecture

File Source

• https://github.com/jiehwan94/AWS_Deployment_House_Price_Prediction_Web_App