Glossary of Common Terms#

The glossary below defines common terms and API elements used throughout aeon.

Channel#
Channels#

A channel is a singular time series in a data set which contains multiple time series variables. A dataset with multiple channels is multivariate.

Collection transformers#

Time series transformers that take a time series collection as input. While these transformers only accept collections, a wrapper is provided to allow them to be used with singular time series datatypes.

Forecasting#

A Time series machine learning task focused on prediction future values of a time series.

Independent variable#
Independent variables#

The variable(s) that are used to predict the target variable(s) in a learning task. Also referred to as exogenous variables Commonly also known as features and attributes in traditional machine learning settings.

Instances#
Instance#

A member of the set of entities being studied and which an machine learning practitioner wishes to generalize. For example, patients, chemical process runs, machines, countries, etc.

May also be referred to as cases, samples, examples, observations or records depending on the discipline and context.

Multivariate#
Multivariate time series#

A time series with multiple channels. Typically observed for the same observational unit. Multivariate time series is typically used to refer to cases where the series evolve together over time.

An example of time series data with multiple channels is data extracted from a gyroscope sensor, which can produce different time series data for the x, y and z axes of the device.

n_jobs#

A parameter for controlling the number of threads used in estimators. Follows the conventions of scikit-learn.

Panel#
Time series panel#

Common alternative name for time series collection.

random_state#

A parameter for controlling random number generation in estimators and functions. Follows the conventions of scikit-learn.

If int, random_state is the seed used by the random number generator; If RandomState instance, random_state is the random number generator; If None, the random number generator is the RandomState instance used by np.random.

Reduction#

Reduction refers to decomposing a given learning task into simpler tasks that can be composed to create a solution to the original task. In aeon reduction is used to allow one learning task to be adapted as a solution for an alternative task.

Seasonality#

When a time series is affected by seasonal characteristics such as the time of year or the day of the week, it is called a seasonal pattern. The duration of a season is always fixed and known.

Series-to-features transformation#

Time series transformers that take a time series as input and output a set of features (in tabular format for time series collections. An example of this is the extraction of the mean and various other summary statistics from the series.

Series-to-series transformation#

Time series transformers that take a time series as input and output a (different) time series. An example of this is the Discrete Fourier Transform (DFT).

Tabular#

A 2 dimensional data structure where the rows of the matrix represent { term}instances and the columns represent variables. This is the most common data structure used in scikit-learn.

A univariate time series can be formatted in this way, where each variable of being measured for each instance are treated as features and stored as a primitive data type in the 2d data structure. E.g., there are N instances of time series and each has T timepoint, this would yield a matrix with shape (N, T): N rows, T columns.

Target variable#
Target variables#

The variable(s) to be predicted in a learning task using Independent variables, past timepoints of the variable itself, or both. Also referred to as the dependent or endogenous variable(s).

Time series annotation#

A learning task focused on labelling the variables of a time series. This includes the related tasks of outlier detection, anomaly detection, change point detection and segmentation.

Time series classification#

A learning task focused on using the patterns across instances between the time series and a categorical target variable.

Time series clustering#

A learning task focused on discovering groups consisting of instances with similar time series.

Time series collection#
Time series collections#

A datatype which contains multiple instances of time series. These series may be univariate time series or multivariate time series. The time series contained within may be of different lengths, sampled at different frequencies, contain differing timepoints etc.

Also referred to as a panel time series depending on context and discipline.

Time series data#
Time series#
Series#

Data with multiple individual variable measurements with accompanying timepoints which are ordered over time or have an index indicating the position of an observation in the sequence of values.

Time series extrinsic regression#

A learning task focused on using the patterns across instances between the time series and a continuous target variable. The aeon regression module is focused on this type of regression.

Time series forecasting regression#

This learning relates to forecasting reduced to regression through a sliding window. This is the more familiar type of regression in literature.

Time series machine learning#

A general term for using machine learning algorithms to learn predictive models from time series data. aeon is a library for time series machine learning algorithms.

Time series regression#

A learning task focused on using learning patterns from multiple time series and a continuous target variable. There are two related but distinct learning tasks that fall under this category: time series forecasting regression and time series extrinsic regression.

A task focused on finding the most similar candidates to a given time series of length l, called the query. The candidates are extracted from a collection of time series of length equal or superior to l.

Time series transformation#
Time series transformers#

Transformers usually refers to classes in the transformation module of aeon. These classes are used to transform time series data into a different format. This may be to reduce the dimensionality of the data, to extract features from the data, or to transform the data into a different format.

See series-to-series transformation and series-to-features transformation for types of transformer.

Timepoint#
Timepoints#

The point in time that an observation is made for a time series. A time point may represent an exact point in time (a timestamp), a time period (e.g. minutes, hours or days), or simply an index indicating the position of an observation in the sequence of values.

Trend#

When time series show a long-term increase or decrease, this is referred to as a trend. Trends can also be non-linear.

Univariate#
Univariate time series#

A single time series.

Variable#
Variables#

Refers to some measurement of interest. Variables may be singular values (e.g. time-invariant measurements like a patient’s place of birth) or a sequence of multiple values as a time series.

For time series data, multiple variables may be referred to as channels.