PYRREGULAR: A Unified Framework for Irregular Time Series, with Classification Benchmarks

TL;DR

This work tackles irregular time series data by introducing pyrregular, a unified framework that converts heterogeneous irregular data into a common array-based representation and provides a standardized 34-dataset repository for classification benchmarks. It details preprocessing to a long format, conversion to a sparse COO tensor, a densification step for downstream learners, and integration with xarray for interoperability across libraries. An extensive benchmark across 12 classifiers reveals that generalist models like Rocket and BORF/LGBM offer robust, fast performance across irregularity types, while deep learning methods can outperform baselines with task-specific fine-tuning. The framework enhances reproducibility and cross-domain evaluation, highlighting both the promise of specialized irregular-time models and the enduring value of simple, robust baselines for irregular data analysis.

Abstract

Irregular temporal data, characterized by varying recording frequencies, differing observation durations, and missing values, presents significant challenges across fields like mobility, healthcare, and environmental science. Existing research communities often overlook or address these challenges in isolation, leading to fragmented tools and methods. To bridge this gap, we introduce a unified framework, and the first standardized dataset repository for irregular time series classification, built on a common array format to enhance interoperability. This repository comprises 34 datasets on which we benchmark 12 classifier models from diverse domains and communities. This work aims to centralize research efforts and enable a more robust evaluation of irregular temporal data analysis methods.

Figures (21)

• Figure 1: An example of an irregular time series, $X$, comprising two signals $\mathbf{x}_1, \mathbf{x}_2$ with indices $\tilde{\mathbf{t}}_1, \tilde{\mathbf{t}}_2$, and the combined shared index $\tilde{\raisebox{0pt}{$\tilde{\mathbf{t}}$}}$.
• Figure 2: Different kinds of irregularity shown on a multivariate time series with $2$ signals and containing up to $5$ timestamps. Missing values are depicted as faded red if they were expected to be recorded, while they are omitted if they are caused by raggedness.
• Figure 3: A simplified schema of our framework. (left) Data from different sources is preprocessed and represented in our proposed array container (center), which combines xarray with an underlying sparse tensor via a custom accessor and backend. This container can be easily manipulated, plotted, and stored. (right) Finally, it can also be converted into a more common dense representation, which can be used for downstream tasks with any standard time series library.
• Figure 4: Long format to coo tensor conversion process. Each row of the long format is processed to retrieve the absolute position $k$ of a given timestamp $t$. The triplet, instance id ($i=1$), signal id ($j=2$), and timestamp index ($k=7$), is used to populate the sparse coo tensor.
• Figure 5: CD plot for the benchmarked models in terms of F1. Best models to the right. Connected models are statistically tied.

Theorems & Definitions (2)

• Definition 2.1: Time Series Signal
• Definition 2.2: Time Series