Description

Learn how to use neuralforecast for time series forecasting with deep learning. Leverage state-of-the-art deep learning models and quickly experiment with their fastest implementation available.

In this notebook, we show how to:

• use cross-validation for model selection
• use NHITS and the LSTM model
• set early stopping to avoid overfitting
• use the Poisson loss to have positive whole integers (since we are dealing with inventory data)
• hyperparameter optimization

We also show how to evaluate your models with metrics specifically built for inventory data, like cumulative forecast error (CFE) or product in stock (PIS).

Easily extend this notebook to include more models and more hyperparameters to tune and develop the best forecasting solution possible.