Back to Devoured - August 06, 2026
DEVOURED
pyhctsa: Python Toolkit for Highly Comparative Time-Series Analysis (GitHub Repo)

pyhctsa: Python Toolkit for Highly Comparative Time-Series Analysis (GitHub Repo)

Data python GitHub
pyhctsa simplifies complex time-series analysis by automating feature extraction with support for parallel processing and customizable feature sets.
What: pyhctsa is a Python toolkit that computes hundreds of structural and statistical features from single or multi-instance time-series data, returning results in a pandas DataFrame.
Why it matters: It addresses the computational burden of time-series analysis by providing a structured, parallelizable framework that lowers the barrier to extracting high-dimensional features.
Takeaway: Install the tool via `pip install pyhctsa` and instantiate `FeatureCalculator()` to begin batch processing your time-series datasets.
Deep dive
  • Features support for custom subsets via YAML configuration.
  • Uses LocalDistributor to parallelize heavy computation across physical CPU cores.
  • Handles varying lengths of time-series instances within a single extraction pass.
  • Includes wrappers for established scientific libraries like TISEAN and specialized medical analysis code.
  • Incorporates LLM-generated code paths verified by human maintainers.
Decoder
  • Time-series feature extraction: The process of transforming temporal data into descriptive numerical vectors that summarize statistical properties like trends, seasonality, and volatility.
Original article

pyhctsa: Python Toolkit for Highly Comparative Time-Series Analysis

Installation

To install pyhctsa you can call:

pip install pyhctsa

Basic Usage

A FeatureCalculator object must first be instantiated using:

from pyhctsa.calculator import FeatureCalculator
calc = FeatureCalculator()

By default, the FeatureCalculator will initialize the full feature set. If you would like to specify a custom feature set, you can pass the corresponding configuration .YAML file as an argument to the FeatureCalculator:

custom_calc = FeatureCalculator(config_path="subset.yaml")

The number of master operations (callable functions) specified by the .yaml will be displayed for verification e.g., Loaded 700 master operations.

Once a FeatureCalculator has been initialized, you can call the extract method to compute time series features on either a single time-series instance or a list of multiple instances:

from pyhctsa.utils import get_dataset

e1000 = get_dataset()
data = e1000[0] # your data as a list, array, or pandas series
res = calc.extract(data)

Note that each time-series instances does not have to be the same length to compute a vector of features. The results of the extraction will be returned in a pandas dataframe of shape N × F, where N is the number of time-series instances and F is the number of time-series features.

Tutorials

New to pyhctsa? Step-by-step tutorials and example workflows are available in the repository.

Advanced Usage

Calling individual operations

If you would like to run individual operations on your data, you can access the corresponding functions from their respective modules directly. For example, to compute the raw_hrv_meas features on your data, the raw_hrv_meas master operation can be accessed from the medical module:

from pyhctsa.operations.medical import raw_hrv_meas

data = ... # your ArrayLike data
res = raw_hrv_meas(data) # result as either a dictionary or scalar value

Individual operations can only be called directly on individual time-series instances.

Parallel Computing

Time-series feature extraction is computationally intensive. To speed up processing, pyhctsa allows you to distribute the workload across multiple CPU cores on your local machine using the LocalDistributor:

from pyhctsa.distribute import LocalDistributor
from pyhctsa.calculator import FeatureCalculator

# initialize the calculator
calc = FeatureCalculator()

# create a LocalDistributor and specify the number of workers
# it is generally recommended to set n_workers to the number of physical CPU cores
dist = LocalDistributor(n_workers=4)

# pass the distributor to the .extract() method
res = calc.extract(data, distributor=dist)

Licenses

Internal licenses

Code for computing features from time-series data is licensed as GNU General Public License version 3.

External packages and dependencies

While the majority of features in pyhctsa rely on standard Python libraries, a small subset of features require external toolboxes.

The following external time-series analysis code packages are provided with the software (in the toolboxes directory), and are used by our main feature-extraction calculator to compute meaningful structural features from time series:

  • Time-series analysis code developed by Michael Small (unlicensed).
  • Max Little's time-series analysis code (GPL License).
  • TISEAN package for nonlinear time-series analysis, version 3.0.1 (GPL license).

The following codebases have been adapted directly into Python code within pyhctsa, rather than being included as external toolboxes:

  • Danny Kaplan's Code for embedding statistics (GPL license).
  • Histogram code by Rudy Moddemeijer (unlicensed).

AI Usage Disclosure

Portions of this codebase (including tests and function documentation) were refactored and generated with the assistance of Large Language Models (LLMs). All AI-generated contributions have been reviewed and verified by the human maintainers.

Explore all 58 articles in today's digest

Fresh dev news, AI-digested daily. Never miss a trend at devoured.fyi