Will I Have A Snow Day.com (WIHASD) is a website that will calculate your chance of having a snow day using an AI model. The model has learned what weather conditions cause snow days by inspecting decades of data, more than any human could analyze.
The site experienced tremendous growth during winter 2024, serving over 130,000 users who relied on its predictions to anticipate potential snow days in their areas.
Importantly, WIHASD has a mechanism that enables it to improve its accuracy over time. When you make a prediction you can sign up to receive an email the day of the predicted snow day. The email will ask if the model’s prediction was correct. The captured weather data and your label (whether or not you had a snow day) will be added to the database used to train the model, thus giving it more information about the weather conditions and your specific zip code.
Powered by XGBoost, scikit-learn, Materialize.css, SendGrid, and Flask. Scraped school closings and processed 100GB+ of weather data from NOAA to build a snow day dataset. Trained a gradient boosting classifier after extensive data exploration and feature engineering.
The site no longer works because the model has not been updated since 2020. Nevertheless, I learned a lot about classical machine learning, handling large datasets, web scraping, building full-stack websites with Flask, and even a little distributed training with Dask.
GitHub Repository: https://github.com/HHousen/willihaveasnowday
Project link: https://willihaveasnowday.com