This is a work in progress, we are adding results as we generate them.

Welcome to the Multiverse

The Multiverse archive for multivariate time series classification.

The Multiverse is an expanded archive for multivariate time series classification (MTSC), together with supporting code, metadata, and benchmark results. It consolidates datasets from the original UEA MTSC archive, newer MTSC collections, donated standalone datasets, and associated benchmark results into a single open repository.

Leaderboards: Multiverse-core (65 datasets) · Full archive (the paper's 100 datasets) · EEG (26 datasets)

The archive is described in The Multiverse of Time Series Machine Learning: an Archive for Multivariate Time Series Classification (arXiv:2603.20352). If you use the archive, please cite it; CITATION.cff has the full entry.

The wider literature the archive is measured against is reviewed in Benchmark-Driven Multivariate Time Series Classification: The Role of the UEA MTSC Archive, a survey of 142 papers using the UEA archive. The accuracies it collected are in published_results/, and are what our runs are compared against.

The current paper version describes:

• 133 unique MTSC problems
• 147 released datasets when preprocessing variants are included
• a curated 65 dataset subset, Multiverse-core (MV-core), for algorithm benchmarking

Multiverse-core leaderboard

#	Estimator	Accuracy rank	Accuracy	Balanced accuracy	AUROC	F1	Log loss ↓	Sensitivity	Specificity
1	HC2	7.40	0.7938	0.7543	0.8951	0.7378	0.5316	0.7570	0.7987
2	MRHydra	8.77	0.7838	0.7529	0.8074	0.7336	7.7909	0.7609	0.7828
3	RDST	9.18	0.7752	0.7396	0.7968	0.7179	8.1012	0.7289	0.7891
4	RIST	9.82	0.7742	0.7430	0.8677	0.7213	0.6123	0.7469	0.7724
5	CIF	10.16	0.7775	0.7463	0.8843	0.7293	0.6414	0.7521	0.7767
6	DrCIF	10.37	0.7722	0.7402	0.8734	0.7216	0.6413	0.7437	0.7718
7	Arsenal	10.80	0.7709	0.7336	0.8469	0.7145	3.6382	0.7340	0.7792
8	QUANT	10.92	0.7651	0.7362	0.8678	0.7177	0.7263	0.7506	0.7554
9	ROCKET	11.04	0.7701	0.7345	0.7927	0.7134	8.2862	0.7315	0.7781
10	LITETime-MV	11.24	0.7483	0.7260	0.8493	0.6851	1.3395	0.7129	0.7668
11	STSF	11.41	0.7699	0.7447	0.8700	0.7145	0.6742	0.7379	0.7827
12	H-InceptionTime	11.89	0.7379	0.7137	0.8434	0.6837	1.3662	0.7192	0.7389
13	Catch22	12.90	0.7539	0.7224	0.8680	0.7027	0.6972	0.7290	0.7507
14	ConvTran	13.12	0.7446	0.7110	0.8520	0.6846	0.9070	0.7219	0.7356
15	PatchMTSC	13.31	0.7443	0.6959	0.8247	0.6682	0.7905	0.6996	0.7381
16	DisjointCNN	13.64	0.7224	0.6975	0.8239	0.6611	2.1344	0.6814	0.7299
17	STC	13.86	0.7518	0.7106	0.8637	0.6853	0.6428	0.7104	0.7599
18	TSF	14.15	0.7419	0.7138	0.8558	0.6926	0.9914	0.7141	0.7498
19	TDE	14.94	0.7272	0.6843	0.8379	0.6601	0.8450	0.6919	0.7304
20	TS2Vec	15.41	0.7192	0.6790	0.7995	0.6562	0.7869	0.6907	0.7125
21	Summary	16.52	0.6936	0.6614	0.8194	0.6393	0.9505	0.6648	0.6979
22	XCM	16.80	0.6691	0.6359	0.7915	0.5859	2.1699	0.6299	0.6769
23	TimesURL	17.23	0.6975	0.6565	0.7833	0.6206	0.9962	0.6456	0.6998
24	TimesNet	17.41	0.7001	0.6647	0.8209	0.6399	1.2957	0.6804	0.6919
25	Dummy	22.70	0.3748	0.3072	0.5000	0.1836	1.3810	0.3248	0.3774

Average over the 57 Multiverse-core datasets with results for every estimator on every metric, ordered by average accuracy rank. Best in each column in bold.

Rebuilt with python -m multiverse.experiments.tables, which also writes a sortable version with per-metric ranks to results/multiverse/leaderboard.html (preview, since GitHub shows HTML as source). Missing results, and why, are listed on that page.

The same command writes a per-dataset view to results/multiverse/datasets.html (preview), which turns the question around: for each dataset it gives the Dummy floor, the median and best over the other estimators, which estimator was best, how much the best gained over Dummy, and how far apart the estimators were. It is sorted by that gain, so the problems where nothing yet beats the baseline come first.

This repository aims to make it easier to:

• load Multiverse datasets through aeon
• inspect archive metadata and dataset variants
• reproduce baseline benchmark results
• compare against published and recreated results
• contribute new results, metadata, and documentation as the archive evolves

Datasets · Results · Leaderboard · Runtime · Memory · Evaluation · Classifiers · Contributing

Installation

Install the release package from PyPI:

pip install aeon-multiverse

or the development version from GitHub:

pip install git+https://github.com/aeon-toolkit/multiverse.git

At present the safest route is to install from source, since the package is changing rapidly:

git clone https://github.com/aeon-toolkit/multiverse.git
cd multiverse
pip install -e .

This repository depends on aeon and uses the aeon dataset loading interface as the main public API for archive access.

Quick start

Load a dataset

The archive datasets are published on Zenodo, in the tsml community. You do not need to download them by hand: use aeon to fetch a dataset from Zenodo and load it into memory.

from aeon.datasets import load_classification

X, y = load_classification("BasicMotions")
train_X, train_y = load_classification("BasicMotions", split="train")
test_X, test_y = load_classification("BasicMotions", split="test")

print(X.shape)

More info and links to code - docs/datasets.md

Train and test a classifier

from aeon.classification.deep_learning import InceptionTimeClassifier
from multiverse.classification import (
    ConvTranClassifier,
    PatchMTSCClassifier,
    TimesNetClassifier,
)

clf = InceptionTimeClassifier()
clf.fit(X, y)
preds = clf.predict(X)

More info and links to aeon classifiers - docs/classifiers.md Multiverse ported classifiers - multiverse/classification

Compare your results to published results

Load results directly in code

from aeon.classification.deep_learning import InceptionTimeClassifier

Or explore published results explored in this repo - docs/results.md

Run an experiment

Results are generated with tsml_eval, which writes one file per classifier, dataset and resample in the format the tooling in this repository reads:

<results_path>/<classifier>/Predictions/<dataset>/testResample<id>.csv

It is the experiments extra, pip install aeon-multiverse[experiments], since only these scripts need it. Its current release pins aeon<1.2.0 and so cannot yet be installed alongside this package; until the next release, use a checkout of tsml_eval main.

Set the data path, results path, classifiers and datasets at the top of main in one of:

• multiverse/experiments/run_single_dataset.py — one classifier on one dataset, the smallest complete example
• multiverse/experiments/run_benchmark.py — a set of classifiers over Multiverse-core
• multiverse/experiments/run_eeg_bakeoff.py — the same, over the EEG archive

then run it:

python -m multiverse.experiments.run_benchmark

Combinations that already have results are skipped, so an interrupted run can be started again, and a failure is reported without ending the run. Running the full benchmark on one machine takes a very long time; the published results were distributed over a cluster.

Donate your code and published results

Coming soon

---

Repository layout

multiverse/
├── docs/                    # Documentation
├── img/                     # Images used in the documentation
├── results/                 # Benchmark results, one directory per classifier
├── published_results/       # Accuracies reported in the MTSC literature
├── survey/                  # Signpost only: the directory moved to published_results/
└── multiverse/              # Python package
    ├── classification/      # Classifiers not available in aeon
    ├── examples/            # Short runnable examples
    └── experiments/         # Result generation and leaderboard tables