3.

5 Data Transformation and Data Discretization 113

data cleaning and was addressed in Section 3.2.2. Section 3.2.3 on the data cleaning
process also discussed ETL tools, where users specify transformations to correct data
inconsistencies. Attribute construction and aggregation were discussed in Section 3.4
on data reduction. In this section, we therefore concentrate on the latter three strategies.
Discretization techniques can be categorized based on how the discretization is per-
formed, such as whether it uses class information or which direction it proceeds (i.e.,
top-down vs. bottom-up). If the discretization process uses class information, then we
say it is supervised discretization. Otherwise, it is unsupervised. If the process starts by first
finding one or a few points (called split points or cut points) to split the entire attribute
range, and then repeats this recursively on the resulting intervals, it is called top-down
discretization or splitting. This contrasts with bottom-up discretization or merging, which
starts by considering all of the continuous values as potential split-points, removes some
by merging neighborhood values to form intervals, and then recursively applies this
process to the resulting intervals.
Data discretization and concept hierarchy generation are also forms of data reduc-
tion. The raw data are replaced by a smaller number of interval or concept labels. This
simplifies the original data and makes the mining more efficient. The resulting patterns
mined are typically easier to understand. Concept hierarchies are also useful for mining
at multiple abstraction levels.
The rest of this section is organized as follows. First, normalization techniques are
presented in Section 3.5.2. We then describe several techniques for data discretization,
each of which can be used to generate concept hierarchies for numeric attributes. The
techniques include binning (Section 3.5.3) and histogram analysis (Section 3.5.4), as
well as cluster analysis, decision tree analysis, and correlation analysis (Section 3.5.5).
Finally, Section 3.5.6 describes the automatic generation of concept hierarchies for
nominal data.

3.5.2 Data Transformation by Normalization

The measurement unit used can affect the data analysis. For example, changing mea-
surement units from meters to inches for height, or from kilograms to pounds for weight,
may lead to very different results. In general, expressing an attribute in smaller units will
lead to a larger range for that attribute, and thus tend to give such an attribute greater
effect or weight. To help avoid dependence on the choice of measurement units, the
data should be normalized or standardized. This involves transforming the data to fall
within a smaller or common range such as [1, 1] or [0.0, 1.0]. (The terms standardize
and normalize are used interchangeably in data preprocessing, although in statistics, the
latter term also has other connotations.)
Normalizing the data attempts to give all attributes an equal weight. Normaliza-
tion is particularly useful for classification algorithms involving neural networks or
distance measurements such as nearest-neighbor classification and clustering. If using
the neural network backpropagation algorithm for classification mining (Chapter 9),
normalizing the input values for each attribute measured in the training tuples will help
speed up the learning phase. For distance-based methods, normalization helps prevent
114 Chapter 3 Data Preprocessing

attributes with initially large ranges (e.g., income) from outweighing attributes with
initially smaller ranges (e.g., binary attributes). It is also useful when given no prior
knowledge of the data.
There are many methods for data normalization. We study min-max normalization,
z-score normalization, and normalization by decimal scaling. For our discussion, let A be
a numeric attribute with n observed values, v1 , v2 , . . . , vn .
Min-max normalization performs a linear transformation on the original data. Sup-
pose that minA and maxA are the minimum and maximum values of an attribute, A.
Min-max normalization maps a value, vi , of A to vi0 in the range [new minA , new maxA ]
by computing

vi minA
vi0 = (new maxA new minA ) + new minA . (3.8)
maxA minA

Min-max normalization preserves the relationships among the original data values. It
will encounter an out-of-bounds error if a future input case for normalization falls
outside of the original data range for A.

Example 3.4 Min-max normalization. Suppose that the minimum and maximum values for the
attribute income are $12,000 and $98,000, respectively. We would like to map income
to the range [0.0, 1.0]. By min-max normalization, a value of $73,600 for income is
73,600 12,000
transformed to 98,000 12,000 (1.0 0) + 0 = 0.716.

In z-score normalization (or zero-mean normalization), the values for an attribute,

A, are normalized based on the mean (i.e., average) and standard deviation of A. A value,
vi , of A is normalized to vi0 by computing

vi A
vi0 = , (3.9)
A

where A and A are the mean and standard deviation, respectively, of attribute A. The
mean and standard deviation were discussed in Section 2.2, where A = n1 (v1 + v2 + +
vn ) and A is computed as the square root of the variance of A (see Eq. (2.6)). This
method of normalization is useful when the actual minimum and maximum of attribute
A are unknown, or when there are outliers that dominate the min-max normalization.

Example 3.5 z-score normalization. Suppose that the mean and standard deviation of the values for
the attribute income are $54,000 and $16,000, respectively. With z-score normalization,
a value of $73,600 for income is transformed to 73,600 54,000
16,000 = 1.225.

A variation of this z-score normalization replaces the standard deviation of Eq. (3.9)
by the mean absolute deviation of A. The mean absolute deviation of A, denoted sA , is

1
sA = (|v1 A| + |v2 A| + + |vn A|). (3.10)
n
 3.5 Data Transformation and Data Discretization 115

Thus, z-score normalization using the mean absolute deviation is

vi A
vi0 = . (3.11)
sA
The mean absolute deviation, sA , is more robust to outliers than the standard deviation,
A . When computing the mean absolute deviation, the deviations from the mean (i.e.,
|xi x|) are not squared; hence, the effect of outliers is somewhat reduced.
Normalization by decimal scaling normalizes by moving the decimal point of values
of attribute A. The number of decimal points moved depends on the maximum absolute
value of A. A value, vi , of A is normalized to vi0 by computing
vi
vi0 = j , (3.12)
10
where j is the smallest integer such that max(|vi0 |) < 1.

Example 3.6 Decimal scaling. Suppose that the recorded values of A range from 986 to 917. The
maximum absolute value of A is 986. To normalize by decimal scaling, we therefore
divide each value by 1000 (i.e., j = 3) so that 986 normalizes to 0.986 and 917
normalizes to 0.917.

Note that normalization can change the original data quite a bit, especially when
using z-score normalization or decimal scaling. It is also necessary to save the normaliza-
tion parameters (e.g., the mean and standard deviation if using z-score normalization)
so that future data can be normalized in a uniform manner.

3.5.3 Discretization by Binning

Binning is a top-down splitting technique based on a specified number of bins.
Section 3.2.2 discussed binning methods for data smoothing. These methods are also
used as discretization methods for data reduction and concept hierarchy generation. For
example, attribute values can be discretized by applying equal-width or equal-frequency
binning, and then replacing each bin value by the bin mean or median, as in smoothing
by bin means or smoothing by bin medians, respectively. These techniques can be applied
recursively to the resulting partitions to generate concept hierarchies.
Binning does not use class information and is therefore an unsupervised discretiza-
tion technique. It is sensitive to the user-specified number of bins, as well as the presence
of outliers.

3.5.4 Discretization by Histogram Analysis

Like binning, histogram analysis is an unsupervised discretization technique because it
does not use class information. Histograms were introduced in Section 2.2.3. A his-
togram partitions the values of an attribute, A, into disjoint ranges called buckets
or bins.