Find research datasets worth reusing
Search datasets from major research repositories and use ShareScore to quickly assess how well each record supports discovery, access, and reuse.
13
datasets available to search
ShareScore release 0.7.1
Dataset results
13 results for “multivariate time series”
Multivariate Time Series data of Fatigued and Non-Fatigued Running from Inertial Measurement Units
<p>The data captured came from mounting a single Shimmer3 IMU on the lumbar of 19 recreational runners. The participants were all regular runners and injury free. The study protocol was reviewed and approved by the human research ethics committee at University College Dublin.<br><br>The data was collected in three segments; in the first, the participant completed a 400m run at a comfortable pace; the second segment consisted of a beep test which acted as the fatiguing protocol for this study; and the last segment where the runner was required to complete the 400m run at their comfortable pace, this time in their fatigued state. The beep test requires the runner to continuously run between two points 20m apart following an audio which produces `beeps' indicating when the person should begin running from one end to the other. The test eventually requires the runner to increase their pace as the interval between the `beeps' reduces as the test progresses. The fatiguing protocol ends when the runner is unable to keep up the increase in pace. The runs were all done on an outdoor running track. The sensor captured acceleration, angular velocity and magnetometer data throughout the three stages of the trials at a sampling rate of 256Hz. The data included here are segmented strides from the two 400m runs of each of the 19 participants. The labels on the data represent the participant number and whether it was a fatigued stride ('F') or a not fatigued stride ('NF').<br>The data used from the sensors includes data from the accelerometer in three directions (X, Y, Z) and the gyroscope in three directions (X, Y, Z). The direction of each of the axis is relative to the sensor. Two extra signals, magnitude acceleration and magnitude gyroscope were derived from the component signals and included in the analysis.</p><p>Kindly cite one of the following papers when using this data:</p><p>B. Kathirgamanathan, B. Caulfield and P. Cunningham, "Towards Globalised Models for Exercise Classification using Inertial Measurement Units," 2023 IEEE 19th International Conference on Body Sensor Networks (BSN), Boston, MA, USA, 2023, pp. 1–4, doi: 10.1109/BSN58485.2023.10331612</p><p>B. Kathirgamanathan, T. Nguyen, G. Ifrim, B. Caulfield, P. Cunningham. Explaining Fatigue in Runners using Time Series Analysis on Wearable Sensor Data, XKDD 2023: 5th International Workshop on eXplainable Knowledge Discovery in Data Mining, ECML PKDD, 2023, <a href="http://xkdd2023.isti.cnr.it/papers/223.pdf">http://xkdd2023.isti.cnr.it/papers/223.pdf</a></p>
Dynamical System Multivariate Time Series
<p>The Dynamical System Multivariate Time Series (DSMTS) Dataset consists of commands, external stimuli, and telemetry readings of a simulated complex dynamical system under fully nominal conditions (no outliers or anomalies).</p> <p>The DSMTS Dataset exhibits a set of desirable properties that make it <strong>very suitable for benchmarking Multivariate Time Series Forecasting </strong>especially for industrial processes of complex systems:</p> <ul> <li><strong>Multivariate (17 variables)</strong> including sensors reading and control signals. It simulates the operational behaviour of an arbitrary complex system including: <ul> <li><strong>4 Deliberate Actuations / Control Commands sent by a simulated operator / controller</strong>, for instance, commands of an operator to turn ON/OFF some equipment.</li> <li><strong>3 Environmental Stimuli / External Forces</strong> acting on the system and affecting its behaviour, for instance, the wind affecting the orientation of a large ground antenna.</li> <li><strong>10 Telemetry Readings representing the observable states of the complex system by means of sensors</strong>, for instance, a position, a temperature, a pressure, a voltage, current, humidity, velocity, acceleration, etc.</li> </ul> </li> <li><strong>5 million timestamps</strong>. Sensors readings are at 1Hz sampling frequency. </li> <li><strong>Pure signal ideal for robustness-to-noise analysis.</strong> The simulated signals are provided without noise: while this may seem unrealistic at first, it is an advantage since users of the dataset can decide to add on top of the provided series any type of noise and choose an amplitude. This makes it well suited to test how sensitive and robust detection algorithms are against various levels of noise.</li> <li><strong>No missing data</strong>. You can drop whatever data you want to assess the impact of missing values on your detector with respect to a clean baseline.</li> </ul>
Accelerometer-Based Multivariate Time-Series Dataset for Calf Behavior Classification
<p><strong>AcTBeCalf Dataset Description</strong></p> <p>The AcTBeCalf dataset is a comprehensive dataset designed to support the classification of pre-weaned calf behaviors from accelerometer data. It contains detailed accelerometer readings aligned with annotated behaviors, providing a valuable resource for research in multivariate time-series classification and animal behavior analysis. The dataset includes accelerometer data collected from 30 pre-weaned Holstein Friesian and Jersey calves, housed in group pens at the Teagasc Moorepark Research Farm, Ireland. Each calf was equipped with a 3D accelerometer sensor (AX3, Axivity Ltd, Newcastle, UK) sampling at 25 Hz and attached to a neck collar from one week of birth over 13 weeks.</p> <p>This dataset encompasses 27.4 hours of accelerometer data aligned with calf behaviors, including both prominent behaviors like lying, standing, and running, as well as less frequent behaviors such as grooming, social interaction, and abnormal behaviors.</p> <p>The dataset consists of a single CSV file with the following columns:</p> <ul> <li><strong>dateTime</strong>: Timestamp of the accelerometer reading, sampled at 25 Hz.</li> <li><strong>calfid</strong>: Identification number of the calf (1-30).</li> <li><strong>accX</strong>: Accelerometer reading for the X axis (top-bottom direction)*.</li> <li><strong>accY</strong>: Accelerometer reading for the Y axis (backward-forward direction)*.</li> <li><strong>accZ</strong>: Accelerometer reading for the Z axis (left-right direction)*.</li> <li><strong>behavior</strong>: Annotated behavior based on an ethogram of 23 behaviors.</li> <li><strong>segId</strong>: Segment identification number associated with each accelerometer reading/row, representing all readings of the same behavior segment.</li> </ul> <p>* the directions are mentioned in relation to the position of the accelerometer sensor on the calf.</p> <p><strong>Code Files Description</strong></p> <p>The dataset is accompanied by several code files to facilitate the preprocessing and analysis of the accelerometer data and to support the development and evaluation of machine learning models. The main code files included in the dataset repository are:</p> <ol> <li><strong>accelerometer_time_correction.ipynb</strong>: This script corrects the accelerometer time drift, ensuring the alignment of the accelerometer data with the reference time.</li> <li><strong>shake_pattern_detector.py</strong>: This script includes an algorithm to detect shake patterns in the accelerometer signal for aligning the accelerometer time series with reference times.</li> <li><strong>aligning_accelerometer_data_with_annotations.ipynb</strong>: This notebook aligns the accelerometer time series with the annotated behaviors based on timestamps.</li> <li><strong>manual_inspection_ts_validation.ipynb</strong>: This notebook provides a manual inspection process for ensuring the accurate alignment of the accelerometer data with the annotated behaviors.</li> <li><strong>additional_ts_generation.ipynb</strong>: This notebook generates additional time-series data from the original X, Y, and Z accelerometer readings, including Magnitude, ODBA (Overall Dynamic Body Acceleration), VeDBA (Vectorial Dynamic Body Acceleration), pitch, and roll.</li> <li><strong>genSplit.py: </strong>This script provides the logic used for the generalized subject separation for machine learning model training, validation and testing.</li> <li><strong>active_inactive_classification.ipynb</strong>: This notebook details the process of classifying behaviors into active and inactive categories using a RandomForest model, achieving a balanced accuracy of 92%.</li> <li><strong>four_behv_classification.ipynb</strong>: This notebook employs the mini-ROCKET feature derivation mechanism and a RidgeClassifierCV to classify behaviors into four categories: drinking milk, lying, running, and other, achieving a balanced accuracy of 84%.</li> </ol> <p>Kindly cite one of the following papers when using this data:</p> <p>Dissanayake, O., McPherson, S. E., Allyndrée, J., Kennedy, E., Cunningham, P., & Riaboff, L. (2024). <em>Evaluating ROCKET and Catch22 features for calf behaviour classification from accelerometer data using Machine Learning models</em>. arXiv preprint arXiv:2404.18159.</p> <p>Dissanayake, O., McPherson, S. E., Allyndrée, J., Kennedy, E., Cunningham, P., & Riaboff, L. (2024). <em>Development of a digital tool for monitoring the behaviour of pre-weaned calves using accelerometer neck-collars</em>. arXiv preprint arXiv:2406.17352</p>
Aeroelastic simulations of wind turbines affected by leading edge erosion: datasets for multivariate time-series classification
<p>This repository contains data generated and used for classification in the publication:<br> Duthé, G.; Abdallah, I.; Barber, S.; Chatzi, E. Modeling and Monitoring Erosion of the Leading Edge of Wind Turbine Blades. <em>Energies</em> <strong>2021</strong>, <em>14</em>, 7262. https://doi.org/10.3390/en14217262</p> <p>The data is generated via OpenFAST aeroelastic simulations coupled with a Non-Homogeneous Compound Poisson Process for degradation modelling and was used to train a Transformer deep learning model.</p> <p>One degradation run generates 1200 samples (1 sample every 6 days corresponding to a 20 year degradation period). In total 20 degradation runs are made available (20x1200 = 24'000 multivariate time-series samples). This repo can serve to benchmark long multivariate time-series classification algorithms. There are 10 possible classes of erosion severity.</p> <p>Each sample is a multivariate time-series of length 60'000, with the following 4 channels extracted from the simulations for a section at the tip of the blade:</p> <ul> <li>Inflow velocity</li> <li>Angle of attack</li> <li>Lift coefficient</li> <li>Drag coefficient</li> </ul> <p>Please see the publication above for more information as well as the included readme for information about the data and an example of how to load it into to PyTorch.</p> <p> </p>
Library of multivariate time series
<p>A database of many different types of multivariate time series, each with between 5-25 processes and between 100-2500 observations.</p> <p>The database contains a serialized Python dictionary of 1053 datasets, where the key for the dictionary is the dataset name, and each value is another dictionary of: "data", an MxT numpy array of processes-by-observations; and "labels", a list of descriptive labels for the dataset.</p>
FrenchPiezo: the French mainland groundwater level multivariate time series
<p><strong>FrenchPiezo </strong>is the French mainland multivariate time series dataset of groundwater levels (a.k.a piezometric levels) across the French mainland. The dataset contains 1026 multivariate time series, each made of 3 dimensions which are:</p> <ul> <li>the piezometric level (<strong>p</strong>),</li> <li>the precipitation (<strong>t</strong><strong>p</strong>),</li> <li>the evapotranspiration (<strong>e</strong>).</li> </ul> <p>Each time series is identified by a code (<strong>bss</strong>) which is the identifier of the piezometer used to measure the piezometric level. The measurements are sampled daily from January 2015 to January 2021, corresponding to 2221 days. The piezometric levels are collected from <a href="https://hubeau.eaufrance.fr/">Hub'Eau</a>, the French free API for accessing water data. The climate data <strong>tp </strong>and <strong>e </strong>are collected from the Copernicus ERA5 archive using their free API. These data are in the file <em>dataset_2015_2021.csv</em><strong> </strong>in which time series with more than 50 missing values have been dropped. The file <em>dataset_2015_2021_nomissing_linear.csv</em> is the same dataset in which missing values have been imputed using linear interpolation.</p> <p>In addition to the climate data, information about the nature of the soil where each time series is collected are downloaded from <a href="https://bdlisa.eaufrance.fr/decouvrir-la-bdlisa"><em>BDLISA</em></a> and are given in the file <em>dataset_stations.csv </em>. BDLISA classifies soils as hydrologic entities (EH), characterized by a set of attributes described <a href="https://www.sandre.eaufrance.fr/ftp/documents/fr/ddd/saq/2.1/sandre_dictionnaire_SAQ__2.1.pdf">here</a>: </p> <p>More information on how we created the dataset, the scripts used to collect the data, and some experiments on forecasting the future values of groundwater levels using global and local foresting methods can be found on our <a href="https://github.com/frankl1/piezoforecast">GitHub page</a>.</p>
Multivariate time series for testing -- RacketSports dataset
<p>The original data was retrieved from http://www.timeseriesclassification.com/description.php?Dataset=RacketSports</p> <p>Original data description:<br> The data was created by university students plyaing badminton or squash whilst wearing a smart watch (Sony Smart watch 35). The watch relayed the x-y-z coordinates for<br> both the gyroscope and accelerometer to an android phone (One Plus 56). The phone<br> wrote these values to an Attribute-Relation File Format (arff) file using an app developed<br> by a UEA computer science masters student. The problem is to identify which sport and which stroke the players are making. The data was collected at a rate of 10 HZ over 3 seconds whilst the player played<br> either a forehand/backhand in squash or a clear/smash in badminton.<br> The data was collected as part of an undergraduate project by Phillip Perks in 2017/18.</p> <p>Pre-processing<br> Data processing was done as described in: https://github.com/NLeSC/mcfly-tutorial/blob/master/utils/tutorial_racketsports.py<br> The original data was split into train and test set. Here the data was loaded and further divided into train, test, validation sets.<br> To keep it simple we here simply divided the original test part into test and validation.<br> The resulting data was stored as numpy .npy files.</p> <p>The zip file contains three sets of time series data (X_train, X_test, X_valid) and the respective labels (y_train, y_test, y_valid).</p> <p>Reference:<br> http://www.timeseriesclassification.com/description.php?Dataset=RacketSports<br> (The data was collected as part of an undergraduate project by Phillip Perks in 2017/18.)</p>
Identifying stationary phases in multivariate time series for highlighting behavioural modes and home range settlements
Open the record for dataset details and reuse information.
FrenchPiezo: the French mainland groundwater level multivariate time series
<p><strong>THIS REPOSITORY IS INCOMPLETE, THE CORRECT DATASET IS AT: <a href="https://zenodo.org/record/7193812#.Y3-ThhTMLic">FrenchPiezo| 7193812#.Y3-ThhTMLic</a> </strong></p> <p>This dataset is a multivariate time series of groundwater level (a.k.a piezometric level) measured by sensors in many cities in the French mainland from January 2015 to January 2021 (2,221 days). The dataset contains 1026 multivariate time series composed of three dimensions sampled daily:</p> <ul> <li><strong>p: </strong>groundwater level</li> <li><strong>tp: </strong>precipitation</li> <li><strong>e: </strong>evapotranspiration</li> </ul> <p>Each time series is identified by a <strong>bss code </strong> which is the identifier of the piezometer using to collect the associated groundwater level.</p> <p>The groundwater level is collected from <a href="https://hubeau.eaufrance.fr/">Hub'Eau</a>, the french service for accessing water data. The precipitation and evapotranspiration are collected from the Copernicus ERA5 climate database. </p> <p>The file <em>dataset_2015_2021.csv</em> contains the raw dataset with missing values</p> <p>The file <em>dataset_nomissing_linear.csv</em> is similar to the previous one, but the missing data have been imputed using linear interpolation.</p> <p> </p> <p>More details about this dataset with source code for collecting it can be found in our paper: </p> <p>```</p> <p>Mbouopda, Michael Franklin, et al. "Experimental study of time series forecasting methods for groundwater level prediction." <em>ECML/PKDD Workshop on Advanced Analytics and Learning on Temporal Data</em>. 2022.</p> <p>```</p>
A Large-Scale Annotated Multivariate Time Series Aviation Maintenance Dataset from the NGAFID
<p>Aviation flight records and maintenance labels. </p>
Artifact for Counterfactual Explanations for Machine Learning on Multivariate HPC Time Series Data
<p>This includes the data sets used in the SC'20 submission "Counterfactual Explanations for Machine Learning on Multivariate HPC Time Series Data".</p>
Multivariate Time Series Search
Multivariate Time-Series (MTS) are ubiquitous, and are generated in areas as disparate as sensor recordings in aerospace systems, music and video streams, medical monitoring, and financial systems. Domain experts are often interested in searching for interesting multivariate patterns from these MTS databases which can contain up to several gigabytes of data. Surprisingly, research on MTS search is very limited. Most existing work only supports queries with the same length of data, or queries on a fixed set of variables. In this paper, we propose an efficient and flexible subsequence search framework for massive MTS databases, that, for the first time, enables querying on any subset of variables with arbitrary time delays between them. We propose two provably correct algorithms to solve this problem — (1) an R-tree Based Search (RBS) which uses Minimum Bounding Rectangles (MBR) to organize the subsequences, and (2) a List Based Search (LBS) algorithm which uses sorted lists for indexing. We demonstrate the performance of these algorithms using two large MTS databases from the aviation domain, each containing several millions of observations. Both these tests show that our algorithms have very high prune rates (>95%) thus needing actual disk access for only less than 5% of the observations. To the best of our knowledge, this is the first flexible MTS search algorithm capable of subsequence search on any subset of variables. Moreover, MTS subsequence search has never been attempted on datasets of the size we have used in this paper.
Fast and Flexible Multivariate Time Series Subsequence Search
Multivariate Time-Series (MTS) are ubiquitous, and are generated in areas as disparate as sensor recordings in aerospace systems, music and video streams, medical monitoring, and financial systems. Domain experts are often interested in searching for interesting multivariate patterns from these MTS databases which can contain up to several gigabytes of data. Surprisingly, research on MTS search is very limited. Most existing work only supports queries with the same length of data, or queries on a fixed set of variables. In this paper, we propose an efficient and flexible subsequence search framework for massive MTS databases, that, for the first time, enables querying on any subset of variables with arbitrary time delays between them. We propose two provably correct algorithms to solve this problem — (1) an R-tree Based Search (RBS) which uses Minimum Bounding Rectangles (MBR) to organize the subsequences, and (2) a List Based Search (LBS) algorithm which uses sorted lists for indexing. We demonstrate the performance of these algorithms using two large MTS databases from the aviation domain, each containing several millions of observations. Both these tests show that our algorithms have very high prune rates (>95%) thus needing actual disk access for only less than 5% of the observations. To the best of our knowledge, this is the first flexible MTS search algorithm capable of subsequence search on any subset of variables. Moreover, MTS subsequence search has never been attempted on datasets of the size we have used in this paper.
ScienceDex guides
Understand access before you commit
These curated guides explain access requirements, typical timelines, costs, and reuse considerations for widely used research datasets.
Allen Brain Atlas
Allen Brain Atlas is an Allen Institute collection of brain map atlases, datasets, APIs, and analysis tools covering mouse, human, and non-human primate brain resources.
Annotated Behaviour and Observability Dataset (ABODe)
ABODe is a University of Edinburgh DataShare dataset for behavior classification in group-housed mice using home-cage video, identities, bounding boxes, ground-plate positions, and annotator labels.
DANDI Archive for NWB datasets
DANDI is a BRAIN Initiative archive for publishing and sharing neurophysiology data, including electrophysiology, optophysiology, and behavioral data packaged as NWB and related standards.
International Brain Laboratory public data
The International Brain Laboratory public data releases expose standardized mouse decision-making experiments, including Neuropixels recordings, widefield calcium imaging, behavior, and session metadata accessed through the ONE API.
OpenNeuro
OpenNeuro is a free, open platform for sharing neuroimaging datasets, with public search, dataset pages, and download paths for web, S3, DataLad, and the OpenNeuro CLI.