TODO:

Calculate WRMSSE

• Test WRMSSE with calculated weights

  • We need to calculate weights using sales via the calculate_weights function
  • Merge it with dummy_preds and test assert

• Then import all of these into Kaggle and calculate naive forecas

Calculate WRMSSE for each hierarchy

• there are 12 hierarchies in total (as defined in definitions.py)
• once we've calculate forecast for each id, we need to aggregate them for specified hierarchies K (1<=K<=12)
• then calculate WRMSEE for for the K hierarchies as defined in the competition guide
• we should output WRMSEE over all hierarchies and also for each hierarchy

Cross Validation

• Then we can look at walk_forward/ sliding window cross validation

Kaggle TO:DO:

• Perform Naive Forecaster in notebook and evaluate with WRMSSE for aggregation level 12