Time Series for Python with PyFlux
summary
PyFlux is a Python library that implements a broad set of time‑series models and inference methods: PyFlux, score-driven (GAS) models and variational state‑space models with options like black box variational inference, Metropolis and MCMC. Examples include a 3,000‑game NFL power‑rating model with team and quarterback random‑walk components and finance examples showing Kalman versus robust t‑GAS filters (robustness to tail shocks). Limitations: no clean smooth estimates for GAS, approximation theory is unsettled, implementation mainly in numpy/pandas with some Cython.
AI-written summary · about
chapters (13)
- 0:00Introduction
- 2:28Time series fundamentals
- 4:53ARIMA models limits
- 7:44State space models
- 13:18Dynamic regression finance
- 16:05Linear Gaussian limits
- 17:50Score-driven models intro
- 21:19Sports analytics
- 24:53NFL team abilities
- 29:31Quarterback effects
- 34:05Model validation
- 39:11PyFlux future
- 41:57Q&A
related talks
28mTime series data science problems with InfluxDB
30mMethods & Tools For Time Series Data Science Problems With InfluxDB
26mTime Series Analysis in Python: Easy Tools for Scientific Insight
29mTime Series Forecasting Techniques in Python
19mForecasting for Everyone - Foundation Models and the Future of Time Series Prediction
01h 43m(Serious) Time for Time Series
42mFlying Python at 40,000 feet
44mMethods and Tools for Time Series Data Science Problems