What are alternatives to PCA for time series data?

Question

I have some data (20 stock price time series) and want to compare different approaches for dimensionality reduction other than PCA (I want to fit only 2 variables in my AR model). I've tried autoencoders, but their reproduction error is very high. Apart from kernel PCA what are other methods I can try?

Answer 1

Besides PCA, Kernel PCA, and Autoencoders you can try:

1. Laplacian Eigenmap (LE)
2. Locally Linear Embedding (LLE)
3. Isometric Mapping (Isomap)
4. Singular Value Decomposition (SVD)
5. Maximum Variance Unfolding (MVU)
6. Locality preserving projection (LPP)
7. Diffusion map (DM)
8. Discrete Fourier Transform (DFT)
9. Discrete Wavelet Transform (DWT)

For more information see:

• M. Ashraf et al., "A Survey on Dimensionality Reduction Techniques for Time-Series Data," in IEEE Access, vol. 11, pp. 42909-42923, 2023, doi: 10.1109/ACCESS.2023.3269693.

A couple more that the authors did not mention are:

10. t-Distributed Stochastic Neighbor Embedding (t-SNE)
11. Symbolic Aggregate approXimation (SAX)