Reference Work Entry

Encyclopedia of Database Systems

pp 2044-2045

# Parametric Data Reduction Techniques

• Rui ZhangAffiliated withUniversity of Melbourne, Melbourne

## Definition

A parametric data reduction technique is a data reduction technique that assumes a certain model for the data. The model contains some parameters and the technique fits the data into the model to determine the parameters. Then data reduction can be performed.

## Key Points

Parametric data reduction (PDR) techniques is opposite to nonparametric data reduction (NDR) techniques. A model with parameters is used in a PDR technique and therefore some computation is required to determine these parameters, which may be costly. However, if a PDR technique is well-chosen, it may result in much more data reduction than NDR techniques. A representative example is linear regression [3]. Linear regression assumes that the data fall on a straight line, expressed by the following formula
$$Y = \,a + bX$$
(1)
Given a set of points (Assuming two dimensions.) {〈x1, y1〉, 〈x2,y2〉,...}, parameters a and b in Equation (1) are determined from the points using the least squares criteria. ...
This is an excerpt from the content