Probabilistic Forecasting with DeepAR and AWS SageMaker

Explore probabilistic forecasting using DeepAR and AWS SageMaker in this 31-minute EuroPython Conference talk. Delve into the theoretical foundations of DeepAR, a deep learning-based algorithm that combines multiple time series for more accurate predictions. Learn how to implement probabilistic forecasting for applications such as energy production, customer demand, and product pricing. Examine a practical time series example and gain hands-on experience with AWS SageMaker implementation. Discover the advantages of DeepAR, including automatic feature engineering and the ability to train on multiple related time series simultaneously. By the end of the talk, acquire the knowledge needed to begin your own forecasting projects using this powerful technique.

Intro
Welcome
Do we need another forecasting algorithm
Probabilistic forecasting
Automatic feature engineering
Multiple time series training
Disadvantages
How it works
Energy Consumption
AWS SageMaker
Prepare Data
Hyper Parameters
Training
Fitting
Questions