Time Series Classification

SHAPES is a dataset of synthetic images designed to benchmark systems for understanding of spatial and logical relations among multiple objects. The dataset consists of complex questions about arrangements of colored shapes. The questions are built around compositions of concepts and relations, e.g. Is there a red shape above a circle? or Is a red shape blue?. Questions contain between two and four attributes, object types, or relationships. There are 244 questions and 15,616 images in total, with all questions having a yes and no answer (and corresponding supporting image). This eliminates the risk of learning biases. Each image is a 30×30 RGB image depicting a 3×3 grid of objects. Each object is characterized by shape (circle, square, triangle), colour (red, green, blue) and size (small, big). Source: Visual Question Answering: A Survey of Methods and Datasets Image Source: https://github.com/ronghanghu/n2nmn#train-and-evaluate-on-the-shapes-dataset

UCI Machine Learning Repository is a collection of over 550 datasets.

The UCR Time Series Archive - introduced in 2002, has become an important resource in the time series data mining community, with at least one thousand published papers making use of at least one data set from the archive. The original incarnation of the archive had sixteen data sets but since that time, it has gone through periodic expansions. The last expansion took place in the summer of 2015 when the archive grew from 45 to 85 data sets. This paper introduces and will focus on the new data expansion from 85 to 128 data sets. Beyond expanding this valuable resource, this paper offers pragmatic advice to anyone who may wish to evaluate a new algorithm on the archive. Finally, this paper makes a novel and yet actionable claim: of the hundreds of papers that show an improvement over the standard baseline (1-nearest neighbor classification), a large fraction may be misattributing the reasons for their improvement. Moreover, they may have been able to achieve the same improvement with a much simpler modification, requiring just a single line of code.

The PhysioNet Challenge 2012 dataset is publicly available and contains the de-identified records of 8000 patients in Intensive Care Units (ICU). Each record consists of roughly 48 hours of multivariate time series data with up to 37 features recorded at various times from the patients during their stay such as respiratory rate, glucose etc. Source: Multi-resolution Networks For Flexible Irregular Time Series Modeling (Multi-FIT) Image Source: https://physionet.org/content/challenge-2016/1.0.0/

Caenorhabditis elegans is a roundworm commonly used as a model organism in the study of genetics. The movement of these worms is known to be a useful indicator for understanding behavioural genetics. Brown {\em et al.}[1] describe a system for recording the motion of worms on an agar plate and measuring a range of human-defined features[2]. It has been shown that the space of shapes Caenorhabditis elegans adopts on an agar plate can be represented by combinations of six base shapes, or eigenworms. Once the worm outline is extracted, each frame of worm motion can be captured by six scalars representing the amplitudes along each dimension when the shape is projected onto the six eigenworms. Using data collected for the work described in[1], we address the problem of classifying individual worms as wild-type or mutant based on the time series. The data were extracted from the C. elegans behavioural database [3]. We have 259 cases, which we split 131 train and 128 test. We have truncated each series to the shortest usable. Each series has 17984 observations. Each worm is classified as either wild-type (the N2 reference strain) or one of four mutant types: goa-1; unc-1; unc-38 and unc-63. [1] A. Brown, E. Yemini, L. Grundy, T. Jucikas, and W. Schafer, A dictionary of behavioral motifs reveals clusters of genes affecting caenorhabditis elegans locomotion, Proceedings of the National Academy of Sciences of the United States of America (PNAS), vol. 10, no. 2, pp. 791 796, 2013. [2] E. Yemini, T. Jucikas, L. Grundy, A. Brown, and W. Schafer, A database of caenorhabditis elegans behavioral phenotypes, Nature Methods, vol. 10, pp. 877 879, 2013. [3] C. elegans behavioural database

This dataset is composed of two collections of heartbeat signals derived from two famous PhysioNet datasets in heartbeat classification, the MIT-BIH Arrhythmia Dataset and the PTB Diagnostic ECG Database. The number of samples in both collections is large enough for training a deep neural network. This dataset has been used in exploring heartbeat classification using deep neural network architectures, and observing some of the capabilities of transfer learning on it. The signals correspond to electrocardiogram (ECG) shapes of heartbeats for the normal case and the cases affected by different arrhythmias and myocardial infarction. These signals are preprocessed and segmented, with each segment corresponding to a heartbeat.

The PLAsTiCC dataset is a collection of simulated light curves from the Photometric LSST Astronomical Time-Series Classification Challenge. This diverse dataset contains 14 types of astronomical time-varying objects, simulated using the expected instrument characteristics and survey strategy of the upcoming Legacy Survey of Space and Time [LSST 79] conducted at the Vera C. Rubin Observatory. The dataset includes two overall categories of time-series objects: transients, which are short-lived events such as supernovae, and variable sources, which are objects with fluctuating brightness such as pulsating stars. Specifically, the dataset includes the following transients: type Ia supernovae (SNIa), SNIax, SNIa-91bg, SNIbc, SNII, superluminous supernovae (SLSN), tidal disruption events (TDE), and single lens microlensing events (µLens-Single); and the following variable objects: active galactic nuclei (AGN), Mira variables, eclipsing binary systems (EB), and RR Lyrae (RRL). The dataset contains a collection of 3,492,890 time-series, providing in each case several fields: time of observation, flux, flux error, and filter in which the flux is measured. Observations in 6 filters are included (u,g,r,i,z,Y). In addition, each time series is associated with a class representing the type of object, and a redshift for that object.

The original dataset for "ECG5000" is a 20-hour long ECG downloaded from Physionet. The name is BIDMC Congestive Heart Failure Database(chfdb) and it is record "chf07". It was originally published in "Goldberger AL, Amaral LAN, Glass L, Hausdorff JM, Ivanov PCh, Mark RG, Mietus JE, Moody GB, Peng C-K, Stanley HE. PhysioBank, PhysioToolkit, and PhysioNet: Components of a New Research Resource for Complex Physiologic Signals. Circulation 101(23)". The data was pre-processed in two steps: (1) extract each heartbeat, (2) make each heartbeat equal length using interpolation. This dataset was originally used in paper "A general framework for never-ending learning from time series streams", DAMI 29(6). After that, 5,000 heartbeats were randomly selected. The patient has severe congestive heart failure and the class values were obtained by automated annotation

The collected dataset consists of multivariate time series (MTS) data belonging to several ATMs banking along with the annotations that the operators did when they performed a maintenance task on any of the machines. Each sample is a MTS with 144 points, associated with all the 10-minute time windows of that day and 38 dimensions. Each dimension is related to a command type and response type, where the value of each point in the time series represents the number of occurrences of the associated command and response since the last failure event., i.e. in this cycle of failure. Therefore, the time series value is accumulated until the next cycle begins. Additionally, some extra information is included for each sample, such as the cycle of failure and the machine identifier, which can be used to create data partitions without mixing different machines. The dataset and the labels assigned to each sample was used in the original work to perform a binary classification problem addressed by ML techniques. The goal of the problem was to predict whether a failure will occur within the next 7 days, using only the information from the current day (accumulated since the last error), which is based on an event-log. Potential use cases of the dataset: • multivariate time series classification-regression-forecasting methodologies; • feature learning- feature extraction approaches; • predictive maintenance tasks: failure classification, failure prediction, anomaly detection.

State-level data for the US economy through the lens of consumer spending (Credit/Debit Spending) . The dataset is enriched with state-level Economic Dynamics and Policy Responses. Specifically, we further enriched the data with the state-level policies as an indication of extreme events (e.g., the state’s business closure order). - Date: The date of the data record. - StateAbbr: Abbreviation of the state. - Population: Population of the state. - Spend_xxx: Various spending metrics (e.g., spend_acf, spend_aer, spend_all). Spend_all (all categories) - Policy: Indicator of policy measures. - lnDailyNewDeaths, DailyNewDeaths: Logarithm and raw counts of daily new deaths. - lnDailyNewCases, DailyNewCases: Logarithm and raw counts of daily new cases. - Deaths, Cases: Cumulative counts. - Day, Day_, DayOfYear: Day-related information.

Recorded with a Husky A200 wheeled UGV, BorealTC contains 116 min of Inertial Measurement Unit (IMU), motor current, and wheel odometry data, focusing on typical boreal forest terrains, notably snow, ice, and silty loam. The dataset also includes experiments on asphalt and flooring. All runs were recorded in Forêt Montmorency and on the main campus of Université Laval, Quebec City, Québec, Canada

The dataset provided is a collection of real-world industrial vibration data collected from a brownfield CNC milling machine. The acceleration has been measured using a tri-axial accelerometer (Bosch CISS Sensor) mounted inside the machine. The X- Y- and Z-axes of the accelerometer have been recorded using a sampling rate equal to 2 kHz. Thereby normal as well as anomalous data have been collected for 4 different timeframes, each lasting 5 months from February 2019 until August 2021 and labelled accordingly. It can be used to investigate the scalability of models and research process variations as the anomaly impact differs. In total there is data from three different CNC milling machines each executing 15 processes. For a detailed description of the data and experimental set-up, please refer to the paper: https://doi.org/10.1016/j.procir.2022.04.022

All data is from one continuous EEG measurement with the Emotiv EEG Neuroheadset. The duration of the measurement was 117 seconds. The eye state was detected via a camera during the EEG measurement and added later manually to the file after analysing the video frames. '1' indicates the eye-closed and '0' the eye-open state. All values are in chronological order with the first measured value at the top of the data.

The data generated from this study are grouped into 3 main types: (1) participant demographic and clinical data, (2) sensor data from the different devices, as well as clinical scores and metadata related to the tasks performed, and (3) participant diaries collected during the in-clinic and at-home phases of the study. Throughout the data tables, timestamps are provided as UNIX epoch/POSIX time.

Dataset contains light curves of 6 rocket body types from Mini Mega Tortora database (MMT)[^1]. The dataset was created to be used as a benchmark for rocket body light curve classification. For more informations follow the original paper: RoBo6: Standardized MMT Light Curve Dataset for Rocket Body Classification[^2] Class labels: - ARIANE 5 R/B - ATLAS 5 CENTAUR R/B - CZ-3B R/B - DELTA 4 R/B - FALCON 9 R/B - H-2A R/B ## Dataset description ## Usage python >>> from datasets import load_dataset >>> dataset = load_dataset("kyselica/RoBo6", data_files={"train": "train.csv", "test": "test.csv"}) >>> dataset DatasetDict({ train: Dataset({ features: ['label', ' id', ' part', ' period', ' mag', ' phase', ' time'], num_rows: 5676 }) test: Dataset({ features: ['label', ' id', ' part', ' period', ' mag', ' phase', ' time'], num_rows: 1404 }) }) - label - class name - id - unique identifier of the light curve from MMT - part - part number of the light curve - period - rotational period of the object - mag - relative path to the magnitude values file - phase - relative path to the phase values file - time - relative path to the time values file Mean and standard deviation of magnitudes are stored in mean_std.csv file. ### File structure - data directory contains 5 subdirectories, one for each class. Light curves are stored in file triplets in the following format: - <track_id>_<#part>_mag.csv - magnitude values - <track_id>_<#part>_time.csv - time values - <track_id>_<#part>_phase.csv - phase angle values where <track_id> is the unique identifier of the light curve from MMT, <\#part> is the part number of the light curve (some light curves are split into multiple parts). - train.csv and test.csv - contains information about the train and test splits (label, id, part, period, mag, phase, time) - mean_std.csv - contains mean and standard deviation for magnitudes, computed over the training set. MMT Rocket Bodies ├── README.md ├── train.csv ├── test.csv ├── mean_std.csv ├── data │ ├── ARIANE 5 R_B │ │ ├── <track_id>_<\#part>_mag.csv │ │ ├── <track_id>_<\#part>_time.csv │ │ ├── <track_id>_<\#part>_phase.csv │ ├── ATLAS 5 CENTAUR R_B │ │ ├── ... │ ├── CZ-3B R_B │ │ ├── ... │ ├── DELTA 4 R_B │ │ ├── ... │ ├── FALCON 9 R_B │ │ ├── ... │ ├── H-2A R_B │ │ ├── ... ## Data preprocessing To create data sutable for both CNN and RNN based models, the light curves were preprocessed in the following way: 1. Split the light curves if the gap between two consecutive measurements is larger than object's rotational period. 2. Split the light curves to have maximum span 1_000 seconds. 3. Filter out light curves which folded form divided into 100 bins has more than 25% of bins empty. 5. Resample the light curves to 10_000 points with step 0.1 seconds. 4. Filter out light curves with less than 100 measurements. ## Citation [^2]: RoBo6: Standardized MMT Light Curve Dataset for Rocket Body Classification @article{kyselica2024robo6, title={RoBo6: Standardized MMT Light Curve Dataset for Rocket Body Classification}, author={Kyselica, Daniel and {\v{S}}uppa, Marek and {\v{S}}ilha, Ji{\v{r}}{\'\i} and {\v{D}}urikovi{\v{c}}, Roman}, journal={arXiv preprint arXiv:2412.00544}, year={2024} } ## References [^1]: Karpov, S., et al. "Mini-Mega-TORTORA wide-field monitoring system with sub-second temporal resolution: first year of operation." Revista Mexicana de Astronomía y Astrofísica 48 (2016): 91-96.

Recorded with a Husky A200 wheeled UGV, the Vulpi 2021 dataset contains 13 min of Inertial Measurement Unit (IMU), motor current, and wheel odometry data, focusing on agricultural terrains. The dataset includes experiments on concrete, a dirt road, a ploughed terrain and an unploughed terrain that were all recorded on an experimental farm in San Cassiano, Lecce, Italy.

Data for the paper entitled Quantifying yeast colony morphologies with feature engineering from time-lapse photography by A. Goldschmidt et al. (https://arxiv.org/abs/2201.05259) This project is a collaboration between Dudley Lab at the Pacific NW Research Institute and the J. Nathan Kutz group at the University of Washington. Summary: Baker's yeast (Saccharomyces cerevisiae) is a model organism for studying the morphology that emerges at the scale of multi-cell colonies. To look at how morphology develops, we collect a dataset of time-lapse photographs of the growth of different strains of S. cerevisiae.

The Tufts fNIRS to Mental Workload (fNIRS2MW) open-access dataset is a new dataset for building machine learning classifiers that can consume a short window (30 seconds) of multivariate fNIRS recordings and predict the mental workload intensity of the user during that window. You can use this dataset for tasks like - time series classification using sliding windows - domain adaptation or domain generalization (how well does your classifier generalize to a new subject?) - fairness of time series classifiers (does performance of your classifier vary by subject race or gender?) Useful Links: * Project Website (and data download links): https://tufts-hci-lab.github.io/code_and_datasets/fNIRS2MW.html * Code for benchmarks: https://github.com/tufts-ml/fNIRS-mental-workload-classifiers * DataSheet documentation: https://github.com/tufts-ml/fNIRS-mental-workload-classifiers/blob/main/Datasheet-Tufts-fNIRS2MW.pdf * Academic Paper (published at NeurIPS Datasets & Benchmarks '21) https://datasets-benchmarks-proceedings.neurips.cc/paper/2021/file/bd686fd640be98efaae0091fa301e613-Paper-round2.pdf ## Motivation We are interested in building brain computer interfaces (BCIs) that would help out everyday computer users working at a desktop or laptop. In our target future use case, a user would actively use a keyboard and mouse as usual, but also wear a non-intrusive headband sensor that would passively provide real-time measurements of brain activity to the computer. Based on moment-to-moment estimates of mental workload, the computer could adjust the interface to support the user. Functional near-infrared spectroscopy (fNIRS) is a promising sensor technology for achieving this goal of "everyday BCI", compared to alternatives like EEG or fMRI. We have developed a prototype fNIRS probe mounted on a headband that we used to collect this dataset (see our paper for details). # Dataset Overview For a complete dataset summary, see our public DataSheet PDF For each participant (68 recommended; 87 total), the dataset contains the following records obtained during one 30-60 minute experimental session. Each subject contributes just over 21 minutes of fNIRS data from the desired n-back experimental conditions, with remaining time related to rest or instruction periods. - fNIRS recordings - Multivariate (D=8) time-series representing brain activity throughout the session, recorded by a sensor probe placed on the forehead and secured via headband - All measurements are recorded at a regular sampling rate of 5.2 Hz. - At each timestep, we record 8 real-valued measurements, one for each combination of - 2 blood chemical concentration changes (oxygenated hemoglobin and deoxygenated hemoglobin) - 2 optical data types used for the measurement (intensity and phase) - 2 spatial locations on the forehead. - The units of each measurement are micro-moles of (oxy-/deoxy-)hemoglobin per liter of tissue. - Activity labels - Annotations of the experimental task activity the subject performed throughout the session, including instruction, rest, and active experiment segments. - We label each segment of the active experiment as one of four possible n-back working memory intensity levels (0-back, 1-back, 2-back, or 3-back). Increased intensity levels are intended to induce an increased level of cognitive workload. - For all experiments reported in the paper, we focus on a binary task (0 vs 2 back) - Demographics - The participant’s age, gender, race, handedness, and other attributes. This lets us measure and audit performance by subpopulation (e.g. how does the classifier perform on white subjects vs. black subjects). ## Publications The Tufts fNIRS Mental Workload Dataset & Benchmark for Brain-Computer Interfaces that Generalize Zhe Huang, Liang Wang, Giles Blaney, Christopher Slaughter, Devon McKeon, Ziyu Zhou, Robert Jacob, and Michael C. Hughes To appear in the Proceedings of Neural Information Processing Systems (NeurIPS 2021) Track on Datasets and Benchmarks , 2021 Link to Paper PDF: <a href="https://openreview.net/pdf?id=QzNHE7QHhut">https://openreview.net/pdf?id=QzNHE7QHhut</a>