Model

A mathematical system that represents or expresses the functional behavior of a real system.

Modeling Data

Also known as "Training Data", this data used by the modeling process to calculate and set model parameters and coefficients. This data is usually highly fit by the models.

Non-Stationary

Relationships that vary and move between data series. Non-stationary relationships hold true for short periods of time.

Optimization

The act of seeking particular actionable variables' values such that one or more results are minimized, maximized or targeted to desired values.

Outlier

An outlier is a data point whose value is significantly infrequent when compared to the frequency of other values.  In uni-modal samples, "infrequent" may be defined as a given number of standard deviations (sigma) from the variable's mean.  In multi-modal systems, an outlier may occur between modes or outside the higher and lower modes.  In multivariate situations, it maybe some defined distance from all cluster centers.

An outlier may be valid or erroneous data.  Particular care should be taken before excluding any outliers unless they can be proven to be erroneous using the underlying knowledge about the data, its collection and it's application.  The decision of inclusion/omission is a difficult one and includes the degree or probability of the data point being erroneous and the impact and risk of negative consequences of omitting each reading individually.

Out-Of-Sample Data

Data held back from the modeling process to determine the true performance of a model.