Real Estate ML Model with Data Scraper

Getting Started

• Must fill-in username in password in data/resources/redfin_login.py
• Create zipcodes.csv in data/resources with headers as [Region, City, Zip Code], Zip Code being the main values for the scraper.
• python run.py to start the webapp. Access through http://localhost:45513/

Predictions

http://localhost:{port}/predictions

• MLS Data is pulled through searching Redfin using the 'Search Redfin' Button. Do not spam or a ban might occur.
  • The app also uses the address to find latitude and longitude (inputs for the model). Searching for the same address breaks this.
• Any current data entered, creates live updates to the predicted sales price at the bottom.
• Current model uses the XGBoost algorithm and has and RMSE of ~45k
• Model is built in ml-model.py and saved as .joblib files for use in the predictions.

Update/Scraper

• To update the data navigate to /data and run python updater.py
• Options will appear to scrape both sales data, MLS data, and Model (combined) data.
• NOTE: Scraping MLS data is very slow (~ 1-2 seconds per address)

Future Plans

• Include more features and switch to neural network
• List Prices would significantly improve the model