neuralforecast 0.0.5

Deep Learning for Time Series Forecasting

DOCUMENTATION

Why NeuralForecast

NeuralForecast is a time-series forecasting library with deep learning models.

Why Deep Learning

• Highly Accurate Predictions:
  • High capacity shared models across panel data time series.
• Fast and Efficient Models:
  • Automatic featurization provided by the networks information processes.
  • Fast GPU computations.

NeuralForecast Features

• Easy-to-use state-of-the-art models:
  • Dataset, dataloader and evaluation utility.
  • Code organization follows Lightning. Pure PyTorch without boilerplate.
  • Implementations of high performing forecasting models with minimal entry barriers.
• High Efficiency and low computation costs:
  • Fast dataloaders and model optimization.
  • Scalable to any hardware without changing the models.

Tutorial 1: Installation and Introduction

Tutorial 2: Time Series DataSets and DataLoaders

Tutorial 3: Model Training and Evaluation

Tutorial 4: Production Deployment

Community

Slack, twitter, something else