Neural Prophet – A powerful AI framework for Time Series Models
summary
Shows Neural Prophet (Python library) using a Torch backend and a PyTorch AR-Net to extend Facebook Prophet’s decomposable model with trend, seasonality, auto‑regression, lagged and future regressors and event windows. Demonstrations train on retail sales and Nifty‑50 daily data, produce a forecast 360 days baseline, add yearly seasonality, then enable n_lags/n_forecast AR‑Net to capture uneven stock volatility better. Live training plots appear; holidays tested but show little impact on this index. Not presented as a trading strategy.
AI-written summary · about
chapters (11)
- 0:00Introduction and setup
- 2:13Time series fundamentals
- 5:38Introduction to NeuralProphet
- 9:08NeuralProphet vs Facebook Prophet
- 11:45Hands-on: Baseline model
- 17:22Forecasting and visualization
- 20:00Advanced modeling with stocks
- 28:20Incorporating seasonality
- 31:41Using AR-Net for volatility
- 33:55Handling recurring events
- 37:06Conclusion and Q&A
related talks
28mBuild Facebook's Prophet in PyMC3
23mComplex AI Forecasting Methods for Investments
33mNixtlaVerse, bridging the gap between statistics and deep learning for time series
30mRealtime Time Series Anomaly Detection in Production
19mForecasting for Everyone - Foundation Models and the Future of Time Series Prediction
38mOptimizing Retail Forecasting Through AI Models
28mNixtla: Deep Learning for Forecasting...
42mSeanTaylorBenLetham