Concise

PyFlux aims to make estimation, inference and forecasting possible in a few lines of code. To achieve this we have a simple API that carries across model types, irrespective of their complexity. This allows you to quickly perform analysis without spending too much time reading documentation.

 

Screenshot from 2016-07-17 17:46:56

 

Modern

PyFlux bringsĀ modern time series models to users. This includes score-driven models, non-Gaussian state space models and modern dynamic volatility models. PyFlux also brings modern methods of inference including black box variational inference that can be used for the time series models on offer.

 

nonlinear

 

Probabilistic

PyFlux time series models treat every quantity as random. Traditionally time series packages have focussed on point estimates, whereas Bayesian packages have usually been aimed at specific models. PyFlux combines breadth of models with breadth of inference. This allows for probabilistic time series modelling.

 

Screenshot from 2016-07-17 17:50:00