Salvador Housing Price Predictor

A Machine Learning API for prediction of housing prices in Salvador/Bahia 📈.

This is an educational personal project for Web Scraping and Machine Learning. All data was scraped from the VivaReal website.

Model features:

areas Total interior habitable area in m²;

bedrooms Total amount of bedrooms. When scraped data contained intervals (e.g. "2-4"), the mean was used, so this feature is of float type;

bathrooms Total amount of bathrooms. When scraped data contained intervals (e.g. "2-4"), the mean was used, so this feature is of float type;

parkingSpots Total amount of parking spots. When scraped data contained intervals (e.g. "2-4"), the mean was used, so this feature is of float type;

type housing type. Can only be "apartment" or "house";

neighborhood neighborhood in Salvador/Bahia in which the housing is located;

neighborhood_area_price value defined by calculating:

(sum of neighborhood's house prices) / (sum of neighborhood's house areas) if the type is "house" or

(sum of neighborhood's apartment prices) / (sum of neighborhood's apartment areas) if the type is "apartment";

prices housing price in BRL;

Installation and usage:

Makefile

If you have make installed, you can use make <target> for interaction with the app.

make init creates the python virtual environment and installs dependencies;

make load_data runs the data Scraping/Cleaning pipeline;

make train_xgb trains and pickles a XGBoost model for the API;

make app runs the API locally and make app_dev runs it on development mode (auto reload). The server will be running on http://127.0.0.1:8000/ and you can check the docs on http://127.0.0.1:8000/docs.

Alternatively, you can run the commands manually:

1. Set virtual environment

Set a Python Virtual Environment. You may use python3 -m venv .venv on your terminal inside the project folder or any other method you prefer.

With your venv created, activate it with source .venv/Scripts/activate on Windows or source .venv/bin/activate on Linux/Mac.

2. Install dependencies

Install the dependencies with pip install -r requirements.txt.

3. Run API

To run the server locally, run uvicorn app:app on the terminal.

And voila, the server will be running on http://127.0.0.1:8000/ and you can check the docs on http://127.0.0.1:8000/docs.

Tools used ⚙️:

• Selenium and undetected_chromedriver
• BeautifulSoup
• Pandas
• Matplotlib and Seaborn
• Scikit-learn
• FastAPI