Forecasting with the Kalman Filter
summary
Introduces the Kalman filter as an implementable forecasting engine using the development statsmodels state space code and a public GitHub notebook, fits local linear trend and unobserved‑components models to industrial electricity data, forecasts out 10 periods and recovers seasonality but finds some regressors (coal price) unhelpful, demonstrates simulation, parameter estimation via maximum likelihood, highlights the Kalman gain, and shows the filter handles missing data and nonstationarity while sometimes yielding wide uncertainty.
AI-written summary · about
chapters (11)
- 0:00Kalman Filter introduction
- 2:52State space models example
- 8:13Simulating and filtering
- 12:56Kalman gain and estimation
- 15:22Forecasting with Kalman
- 17:55Parameter estimation
- 19:41Unobserved components
- 22:11AR models in state space
- 24:01Industrial electricity demand
- 28:39Extensions to Kalman
- 30:25Q&A
related talks
01h 26mApplying Statistical Modeling & Machine Learning to Perform Time-Series Forecasting
37mTracking the tracker: Time Series Analysis in Python From First Principles
29mTime Series Forecasting Techniques in Python
01h 30mTime Series Analysis with StatsModels
47mIPython and Sympy to Develop a Kalman Filter for Multisensor Data Fusion
42mRobust Automated Forecasting in Python and R
40mDeep Learning and Time Series Forecasting for Smarter Energy
32mAdopting Open-Source Tools for Time Series Forecasting