A normalization transformation scales continuous (numerical) values to the range -1.0 to 1.0 or the range 0.0 to 1.0.
A normalization transformation is of the form xnew=((xold-shift)/scale)
. Normalization applies to numerical attributes only.
Oracle Data Miner supports the following normalization schemes:
(x - MIN(x))/(MAX(x) - MIN(x))*(new max - new min) + new min
(x - MEAN(x))/SQRT(VARIANCE(x))
Min-Max recodes all values to be in the range 0.0 to 0.1; the relative position of each value is maintained. You can change the range.
If the data has outlier values that must be preserved, Z-Score normalizes most values to the range from –1 to 1, but allows values outside that range representing the outliers.
(x/MAX(ABS(MIN(x)), ABS(MAX(x))))
. This method normalizes sparse values to values between -1 and +1.If you select Min-Max, you can specify the Min (minimum) and Max (maximum) of the normalized distribution; the defaults are 0 for the minimum and 1 for the maximum.
Copyright © 2006, 2008, Oracle. All rights reserved.