09-19, 13:40–14:30 (Europe/Amsterdam), Escher
We provide an introduction to implementing time series forecasting modeling in NumPyro. This allows us to write custom models for which we can have complete control. This includes demand forecasting estimation via censoring likelihoods and hierarchical time series models.
We provide an introduction to implementing time series forecasting modeling in NumPyro. This allows us to write custom models for which we can have complete control. This includes demand forecasting estimation via censoring likelihoods and hierarchical time series models. NumPyro is a probabilistic programming language relying on JAX. It offers inference methods like NUTS (MCMC) and stochastic variational inference. In this talk, we describe how to implement classical forecasting methods like exponential smoothing, auto-regressive models, and techniques for sparse time series like Croston and TSB. Then, we show how to extend these methods to create customer models by, for example, using non-gaussian innovations (like Student-T and censoring likelihoods) and using hierarchical models.
Bonus: We briefly describe how we can implement similar models in PyMC.
References:
- Notes on Exponential Smoothing with NumPyro
- Croston's Method for Intermittent Time Series Forecasting in NumPyro
- Bayesian Censoring Data Modeling
- Demand Forecasting with Censored Likelihood
- Hierarchical Exponential Smoothing Model
- PyMC Example: Time Series Models Derived From a Generative Graph
- Time Series Analysis with Bayesian State Space Models in PyMC
Juan is a Mathematician (Ph.D. Humboldt Universität zu Berlin) and data scientist. He is interested in interdisciplinary applications of mathematical methods. In particular, time series analysis, bayesian methods, and causal inference.