Glossary

Glossary

It consolidates uncommon or specialized terms used in TIM anomaly detection, is meant to save time and promote consistency.

term	definition
Anomalous behavior model	It is a important part of TIM anomaly detection engine, that is responsible for creating model for each detection perspective resulting in anomaly indicator.
Anomaly	Data point that does not conform to the expected behavior of a given group, i.e. data point that does not fit well with the rest of the data (often referred to as anomaly, outlier, exception or contaminant).
Anomaly indicator	Final output of TIM Anomaly Detection
Autocorrelation	Autocorrelation, also known as serial correlation, is the correlation of a signal with a delayed copy of itself as a function of delay.
Causation	Causation indicates a relationship between two variables where one variable is affected by the other. For example, when the value of one variable, increases or decreases as a result of other variables, it is said there is causation.
Collective anomaly	Type of anomaly
Configuration	All available settings of TIM Anomaly Detection
Contextual anomaly	Type of anomaly
Correlation	Correlation is a statistical measure that reveals the degree to which two variables are linearly related (no declaration of cause and effect).
Dependency-oriented data	Time-series are dependency-oriented data. In such data, anomalies are usually defined in a contextual or collective sense and are harder to distinguish from noise.
Design of experiment	Design of experiment is characterized by data(chosen KPI and influencers affecting it) and domain specifics
Detection	Evaluation of an existing model
Domain specifics	Settings that are related to your domain
Feature	Feature is a transformation(by engine) of the original variable/variables
Global anomaly	Type of anomaly
Influencer	Influencers(explanatory variables) are the ones which through we are trying to explain the value or effect on the KPI variable by creating a relationship between an explanatory and dependent variable.
Key performance indicator(KPI)	KPI(dependent variable) is nothing but the variable of a dataset about which we want to have a deeper understanding - we are detection anomalies on this variable
Model building	Creation of a new model
Model rebuilding	Reconstruction of an existing model
Multivariate AD with defined KPI	More dimensional AD problem, that has to be characterized by one of the columns in the data - KPI
Normal behavior model	It is an important part of TIM anomaly detection engine, that is responsible for creating model characterizing the expected behavior of a given KPI
Normal behavior	The output of normal behavior model representing the expected value of a given KPI
Detection perspective	The aspect of how you look on anomalies.
Rebuild type	Define what part of a model has to be reconstructed
Residuals	A difference between actual and normal behavior values of a given KPI
Root cause analysis	An interpretation of what drives normal behavior
Semi-supervised AD	Type of an algorithm related to label availability
Sensitivity	A number in percentage that defines the sensitivity of the underlying model to anomalies
Supervised AD	Type of an algorithm related to label availability
Temporal continuity	Temporal continuity refers to the fact that the patterns in the data are not expected to change abruptly, unless there are abnormal processes at work
Time-series data	Time series data is a collection of observations for a single subject (entity) obtained through repeated measurements at different time intervals (generally equally spaced as in the case of metrics, or unequally spaced as in the case of events).
Univariate AD	One dimensional AD problem
Unsupervised AD	Type of an algorithm related to label availability
Variable	By variable we mean either KPI(dependent variable) or Influencer(explanatory variable).