Chapter 7 · Part 3
Feature Engineering for Time Series
Twelve sections of feature engineering — statistical, autocorrelation, spectral, shape-based, decomposition-based, embedding, multivariate, holiday, robust, and chaotic-dynamics features. Includes representation learning with TS2Vec and TS-TCC.
Learning objectives
By the end of this chapter you will be able to:
- ▸Build statistical and distributional features from a series
- ▸Decide when frequency-domain features outperform autocorrelation features
- ▸Apply TS2Vec / TS-TCC representation learning to short series
- ▸Engineer holiday, event, and calendar features that hold up under drift
- ▸Use chaotic-dynamics features when classical features fail
- 1.1Introduction to Time Series Feature Engineering
- 2.1Statistical and Distributional Features
- 3.1Temporal and Autocorrelation Features
- 4.1Frequency-Domain and Spectral Features
- 5.1Shape-Based and Pattern Features
- 6.1Decomposition-Based Features
- 7.1Embedding and Representation Features
- 8.1Time Series Representation Learning: TS2Vec and TS-TCC
- 9.1Multivariate and Causality-Aware Features
- 10.1Holiday and Event Calendars
- 11.1Noise-Based and Robust Features
- 12.1Chaotic and Nonlinear Dynamics Features
- 13.1Forecast Horizon Strategies and Feature Engineering
- 14.1Unsupervised and Autoencoder-Based Features
- 15.1Domain-Specific Feature Sets
- 16.1Tooling Ecosystem and Pipeline Design
- 17.1Best Practices for Implementation
- 18.1Techniques by Forecasting Objective
- 19.1Conclusion and Future Directions
- 20.1Reference Implementations and Libraries
Get the full chapter
All eleven chapters · 478 references · private Discord community.
← Previous · Chapter 6
State-of-the-Art Performance Evaluation
Next · Chapter 8 →
Advanced Machine Learning for Forecasting