CRAZY Lots of work
Sorry I really like procrasting so here is whats up:
I worked on connecting the different parts of the project instead of leaving them as a scraper, some models, and a website that do not talk to each other. The project is definitely not finished, but there is now a prototype showing the whole scrape to forecast and predictinions
Scraping
I created a scraper to collect hourly data for a bunch of weather stations. I only used one for now. it uses playwright and bs4
Create ML models to predict the wind
I had LSTMs, ridge (like lr), and naive persistence models predict wind components.
The models now train, calculate MAE, and save their predictions, models, and preprocessing settings.
Scoring
I created a header scorer that takes the predicted winds and ranks example directions based on how much wind would direcly oppose travel (N, E, S, W base(
Ridge Choosing a heading from forcast
Persistance chooses from latest
Both choices are scored using the wind actually observed in the target hour.
Only a early test for early battery savings/
Website
I connected the saved predictions and scorer to my FastAPI/Jinja website.
Replay page lets you choose a historical hour and see the forcast and heading and more. Its not live inference yet
Comments 0
No comments yet. Be the first!
Sign in to join the conversation.