Hi! This project has the purpose to predict the weather (wind’s force and direction) into the Biscay bay in France.
The rough data used for that work are here:
Inside this repository you’ll find a package (meteo_marine_traitement) able to clean, convert, trim and store data ready for sklearn and its notice in english. You’'ll find a notebook too.
The best score i got was with a dataset of all the variables, 3 days per row and a RandomForestClassifier model. I have tried deep learning with keras on it, using a MPC framework but the score is almost the same. Maybe RNN framework would be a good option to improve the score because each row is sequential in time…
To improve it i have found a website with data about energy production from sun panels in France; maybe a little lane to find a link between that and cloud cover…
Here is the adress on github:
advices and comments are welcomed !
(i’ve almost forgotten, if you are interested i have done a similar work on the weather above the north ouest of France, i can translate it in english and display it with a notebook if you want)