Is there a way to select segments from different original time series that can be combined to create a new time series that best represents the information of the original multiple time series?

Question

I am looking for a way to select segments from different original time series that can be combined to create a new time series that best represents the information of the original multiple time series.

Some machine learning method?

Answer 1

Hola Jie,

This is a good question. While it is not possible to reconstruct multiple time series using a subset of the original data, you could consider using Principal Component Analysis (PCA / SVD) to reduce the dimensionality of the data, then use that subset of its principal components to reconstruct the data in a way that closely resembles the original multiple time series data.

Here is "Principal Component Analysis (PCA): A Step-by-Step Explanation" that speaks broadly about PCA -- https://builtin.com/data-science/step-step-explanation-principal-component-analysis -- as well as "PCA Using Python: A Tutorial," a tutorial that runs you through a basic example of using PCA to find the principal components of the well-known "Iris" dataset -- https://builtin.com/machine-learning/pca-in-python --

Look forward to hearing how that goes!