NVE / avalanche_ml

Applying machine learning in the Norwegian Avalanche warning Service
MIT License
2 stars 1 forks source link

Functional aggregation from Varsom and RegObs #3

Closed widforss closed 4 years ago

widforss commented 4 years ago

This module is somewhat complete now. More options are needed fine-tuning what data goes into a LabeledData, however. ForecastDataset should also be changed in the future to support RegObs and weather data for days without an published avalanche bulletin.

The suggested workflow is to create a ForecastDataset once, generate a number of different LabeledData from it and write those to csv-files. To use the datasets later, use LabeledData.from_csv() to read in from disk.

This way, the only recurring delay using the data is to read CSVs from disk, instead of relying on the web services or pickled caches of data formatted in an unsuitable manner.

widforss commented 4 years ago

Had to force-push to get things straight after pushing to widforss/master in the middle of todays outage.