Nonparametric Data Reduction Techniques
A nonparametric data reduction technique is a data reduction technique that does not assume any model for the data.
Nonparametric data reduction (NDR) techniques is opposite to parametric data reduction (PDR) techniques. A PDR technique must assume a certain model for the data. Parameters of the model are determined before the data reduction is performed. A NDR technique does not assume any model and is applied to the data directly. The data reduction effectiveness of a PDR technique heavily depends on whether the model suits the data well. If well-suited, good accuracy as well as substantial data reduction can be achieved; otherwise, both cannot be achieved at the same time. A NDR technique yields more uniform effectiveness irrespective of the data, but it may not achieve as high data reduction as a well-suited PDR technique.
Popular NDR techniques include histograms, clustering and indexes. Histograms are used to approximate data distributions. An equidepth...
- 1.Barbará D, DuMouchel W, Faloutsos C, Haas PJ, Hellerstein JM, Ioannidis YE, Jagadish HV, Johnson T, Ng RT, Poosala V, Ross KA, Sevcik KC. The New Jersey data reduction report. Q Bull IEEE TC Data Eng. 1997;20(4):3–45.Google Scholar