miriamkw/GluPredKit

Add "helpers"

Closed this issue · 1 comments

miriamkw commented

Helper methods/classes to process data:

  • Data wrangler
    • Merging data into the same time grid
    • Removing duplicates
    • Handling missing data (NOT REMOVING NANS! We have ALL the time grids, so after adding lookback etc. we can remove nan)
  • Data augmenter
    • CGM delta
    • Basal, bolus and both combined
    • Adding lookbacks
    • Adding future (known) inputs
    • Adding targets
    • Deleting nans
  • Scikit and pytorch dataset generator
    • Scaling: categoical, numerical, controls, not controls
    • Which inputs to use
    • Hyperparameter tuning
    • scaling
    • pipeline
    • ..?
miriamkw commented
  1. Data Wrangling
    • Merging data
    • Adding time lagged features and targets
    • Removing nan values
  2. Data Splitting
    • Making sure that there is no information leakage
  3. Preparing for machine learning library
    • Scikit learn
      • Scaling and one hot encoding
    • PyTorch
      • Scaling and one hot encoding