AutoAI glossary

Learn terms and concepts that are used in AutoAI for building and deploying machine learning models.

aggregate score
The aggregation of the four anomaly types: level shift, trend, localized extreme, variance. A higher score indicates a stronger score.

algorithm
A formula applied to data to determine optimal ways to solve analytical problems.

AutoAI experiment
An automated training process that considers a series of training definitions and parameters to create a set of ranked pipelines as model candidates.

batch employment
Processes input data from a file, data connection, or connected data in a storage bucket and writes the output to a selected destination.

bias detection (machine learning)
To identify imbalances in the training data or prediction behavior of the model.

binary classification
A classification model with two classes and only assigns samples into one of the two classes.

classification model
A predictive model that predicts data in distinct categories.

confusion matrix
A performance measurement that determines the accuracy between a model’s positive and negative predicted outcomes to positive and negative actual outcomes.

cross validation
A technique that tests the effectiveness of machine learning models. It is also used as a resampling procedure for models with limited data.

data imputation
Substituting missing values in a data set with estimated values.

exogenous features
Features that can influence the prediction model but cannot be influenced in return. See also: Supporting features

fairness
Determines whether a model produces biased outcomes that favor a monitored group over a reference group. Fairness evaluations detect if the model shows a tendency to provide a favorable or preferable outcome more often for one group over another. Typical categories to monitor are age, sex, and race.

feature correlation
The relationship between two features. For example, postal code might have a strong correlation with income in some models.

feature encoding
Transforming categorical values into numerical values.

feature importance
The relative impact a particular column or feature has on the model's prediction or forecast.

feature scaling
Normalizing the range of independent variables or features in a data set.

feature selection
Identifying the columns of data that best support an accurate prediction or score.

feature transformation
In AutoAI, a phase of pipeline creation that applies algorithms to transform and optimize the training data to achieve the best outcome for the model type.

holdout data
Data used to test or validate the model's performance. Holdout data can be a reserved portion of the training data, or it can be a separate file.

hyperparameter optimization (HPO)
The process for setting hyperparameter values to the settings that provide the most accurate model.

incremental learning
The process of training a model that uses data that is continually updated without forgetting data that is obtained from the preceding tasks.

large tabular data
Structured data that exceeds the limit on standard processing and must be processed in batches. See incremental learning.

labeled data
Data that is labeled to identify the appropriate data vectors to be pulled in for model training.

monitored group
A class of data monitored to determine whether the results differ significantly from the results of the reference group. For example, in a credit app, you might monitor applications in a particular age range and compare results to the age range more likely to recieve a positive outcome to evaluate whether there might be bias in the results.

multiclass classification model
A classification task with more than two classes. For example, where a binary classification model predicts yes or no values, a multi-class model predicts yes, no, maybe, or not applicable.

multivariate time series
Time series experiment that contains two or more changing variables. For example, a time series model that forecasts the electricity usage of three clients.

optimized metric
The metric used to measure the performance of the model. For example, accuracy is the typical metric that is used to measure the performance of a binary classification model.

pipeline (model candidate pipeline)
End-to-end outline that illustrates the steos in a workflow.

positive class
The class that is related to your objective function.

reference group
A group that you identify as most likely to receive a positive result in a predictive model. You can then compare the results to a monitored group to look for potential bias in outcomes.

regression model
A model that relates a dependent variable to one or more independent variable.

scoring
In machine learning, the process of measuring the confidence of a predicted outcome.

supporting features
Input features that can influence the prediction target. See also: Exogenus features

text classification
A model that automatically identifies and classifies text into distinct categories.

time series model (AutoAI)
A model that tracks data over time.

trained model
A model that is ready to be deployed.

training
The initial stage of model building, involving a subset of the source data. The model can then be tested against a further, different subset for which the outcome is already known.

training data
Data used to teach and train a model's learning algorithm.

univariate time series
Time series experiment that contains only one changing variable. For example, a time series model that forecasts the temperature has a single prediction column of the temperature.