A service predicting the probability of Snoqualmie Pass Road Closure

Problem: Snoqualmie Pass drivers need to be better prepared for disruption to their winter travel plans

Solution: A service predicting the probability of Snoqualmie Pass Closure based on historical and weather data

At pass-closure.com users are able to see the probability of Snoqualmie Pass Closure over the next 15 days. The website was created using Flask and is hosted on AWS.

• Historic weather data from NOAA using the ASOS (Automated Surface Observing Systems) for Stampede Pass
• Snoqualmie Pass closure data requested from WSDOT
• Forecast weather data scraped from weather.com to feed into model

I used a logistic regression model as my baseline model with basic data of temperature and 1s/0s for: precipitation, overcast, poor-visibility, windy.

I tried other models and features including Random Forest and Gradient Boosting. Then chose a random forest model as my final model with aggregated daily data including year, month, day and get_dummies for day of week as well as the original data.

Clone the repo, then run the following commands:

• from the src folder:
  • python initialize_db.py
  • Note - because this is an hourly process the first data will be populated to the MongoDB after an hour
• from the website folder:
  • export FLASK_APP=web_app.py
  • flask run

• Move to a hourly granularity for at least the first 3 days, with 2 separate models, to give users a better idea of the best time of day to use the pass
• Expand to other frequently used mountain passes

• Washington State Department of Transport (WSDOT)
• Robert Sherer's dressing for the weather project used as a reference for web scraping: repo