Skip to the content
Know your tech before you buy it. Makers: get listed from $49/mo

timeseriesflattener

aarhus-psychiatry-research.github.io

  • Free plan
  • Linux
  • API

What timeseriesflattener is

timeseriesflattener is a Python package that builds time-series features for machine-learning predictors. It turns irregular observations into a dataframe with one row for each prediction time, grouping values by an ID column and placing generated features in columns. Feature specifications can define prediction times, lookbehind windows for predictors, and lookahead windows for outcomes. Documented aggregations include count, earliest, latest, maximum, mean, minimum, slope, sum, unique count, and variance. The API supports temporal predictors, outcomes, boolean outcomes, static features, and time-delta features. A fallback value can cover a look window with no relevant data. The package accepts pandas or Polars dataframes for prediction-time and static inputs. Tutorials show flattened predictors from pre-embedded text with entity IDs, timestamps, and embedding columns. Operations can use n_workers for parallel processing across multiple cores. Installation uses pip, and downloadable Jupyter notebooks can run locally. The documentation directs bugs and feature requests to GitHub Issues and discussion to GitHub Discussions.

Who it suits

It is especially suited to people handling complicated, irregular time series while training simple models. Python users working with pandas or Polars dataframes can use its documented feature specifications.

Good to know, and what to watch

  • Handles irregular time-series observations
  • Supports pandas and Polars dataframes
  • Offers multiple documented aggregation methods
  • Allows predictor and outcome windows
  • Can parallelize operations across cores
  • Requires Python package installation through pip
  • No pricing or trial terms are stated

The verdict

timeseriesflattener focuses on converting irregular time-series data into structured model features. Its window controls, aggregation choices, fallback values, and dataframe support suit feature-engineering workflows.

What the maker publishes

Aggregations
Documented aggregators include count, earliest, latest, maximum, mean, minimum, slope, sum, unique count, and variance. source
Audience
The introductory tutorial says the package is especially helpful for complicated and irregular time series when training simple models. source
Dataframes
The API accepts pandas or Polars dataframes for prediction-time and static frames. source
Feature types
The API documents temporal predictors, outcomes, boolean outcomes, static features, and time-delta features. source
Install
The official installation page instructs users to install the package with pip using `pip install timeseriesflattener`. source
Missing values
Feature specifications accept a fallback value for cases where the relevant look window has no data. source
Output
It converts irregular time series into a dataframe with one row per prediction time and columns for constructed features, aggregating values by an ID column. source
Parallel processing
The introductory tutorial says n_workers can parallelize operations across multiple cores. source
Prediction windows
Feature specifications let users set prediction times and lookbehind windows for predictors or lookahead windows for outcomes. source
Pricing and trial
The opened official documentation describes a Python package and installation instructions but states no price or free-trial terms. source
Publication
The package has a 2023 paper in the Journal of Open Source Software describing it as a Python package for summarizing features from medical time series. source
Purpose
timeseriesflattener is a Python package for generating time-series features used as predictors in machine-learning models. source
Support
The docs direct bug reports and feature requests to GitHub Issues and usage questions or general discussion to GitHub Discussions. source
Text
The text tutorial demonstrates generating flattened predictors from pre-embedded text represented as a dataframe with entity IDs, timestamps, and embedding columns. source
Tutorials
The documentation provides downloadable Jupyter notebook tutorials that users can run locally. source

Checked against the maker's pages on 4 Oct 2026.