Adama Science & Technology University

SoEEC:CSE
Data Mining(CSE4301)
Chapter 3

Data preprocessing

1
1
 Lecture out line

 Why preprocess the data?

 Data cleaning
 Data integration and transformation
 Data reduction
 Discretization and concept hierarchy
generation
 Summary
 Why Data Preprocessing?

 Data in the real world is dirty

 incomplete: lacking attribute values, lacking certain
attributes of interest, or containing only aggregate data
 noisy: containing errors or outliers
 inconsistent: containing discrepancies in codes or
names
 No quality data, no quality mining results!
 Quality decisions must be based on quality data
 Data warehouse needs consistent integration of quality
data
 Required for both OLAP and Data Mining!
 Why can Data be Incomplete?
 Attributes of interest are not available (e.g.,
customer information for sales transaction data)
 Data were not considered important at the time
of transactions, so they were not recorded!
 Data not recorded because of misunderstanding
or malfunctions
 Data may have been recorded and later deleted!
 Missing/unknown values for some data
Why can Data be Noisy/Inconsistent?

 Faulty instruments for data collection

 Human or computer errors
 Errors in data transmission
 Technology limitations (e.g., sensor data come at
a faster rate than they can be processed)
 Inconsistencies in naming conventions or data
codes (e.g., 2/5/2002 could be 2 May 2002 or 5
Feb 2002)
 Duplicate tuples, which were received twice
should also be removed
Major Tasks in Data Preprocessing
 Data cleaning
 Fill in missing values, smooth noisy data, identify or remove
outliers, and resolve inconsistencies
 Data integration
 Integration of multiple databases, data cubes, or files
 Data transformation
 Normalization and aggregation
 Data reduction
 Obtains reduced representation in volume but produces the
same or similar analytical results
 Data discretization
 Part of data reduction but with particular importance,
especially for numerical data
Forms of data preprocessing
 Data Cleaning

 Data cleaning tasks

 Fill in missing values
 Identify outliers and smooth out noisy data
 Correct inconsistent data
 The first step in data cleaning as a process is
discrepancy detection.
 Discrepancies can be caused by several
 poorly designed data entry forms that have many optional
fields, human error in data entry,
 deliberate errors (e.g., respondents not wanting to divulge
information about themselves), and data decay (e.g.,
 outdated addresses).
 Discrepancies may also arise from inconsistent
data representations and inconsistent use of
Codes.
 How to Handle Missing Data?
 Ignore the tuple: usually done when class label is missing
(assuming the tasks in classification)—not effective when the
percentage of missing values per attribute varies considerably.
 Fill in the missing value manually: tedious + infeasible
 Use a global constant to fill in the missing value: e.g.,
“unknown”, a new class!
 Use the attribute mean to fill in the missing value
 Use the attribute mean for all samples belonging to the same
class to fill in the missing value: smarter
 Use the most probable value to fill in the missing value:
inference-based such as Bayesian formula or decision tree
How to Handle Missing Data?
Age Income Position Gender
30 24,200 Manager M
39 ? Assistant F
45 45,390 Professor F
30 30,000 Manager ?
27 22000 ? F

Fill missing values using aggregate functions (e.g., average) or

probabilistic estimates on global value distribution
 Noisy Data

 Noise: random error or variance in a measured

variable
 Incorrect attribute values may exist due to
 faulty data collection instruments
 data entry problems
 data transmission problems
 technology limitation
 inconsistency in naming convention
 Other data problems which requires data cleaning
 duplicate records
 incomplete data
 inconsistent data
 How to Handle Noisy Data?
Smoothing techniques
 Binning method:
 first sort data and partition into (equi-depth) bins
then one can smooth by bin means, smooth by bin median,
smooth by bin boundaries, etc.
 Clustering
 detect and remove outliers
 Combined computer and human inspection
 computer detects suspicious values, which are then
checked by humans
 Regression
 smooth by fitting the data into regression functions
 Use Concept hierarchies
 use concept hierarchies, e.g., price value -> “expensive”
Simple Discretization Methods: Binning
 Equal-width (distance) partitioning:
 It divides the range into N intervals of equal size:
uniform grid
 if A and B are the lowest and highest values of the
attribute, the width of intervals will be: W = (B-A)/N.
 The most straightforward
 But outliers may dominate presentation
 Skewed data is not handled well.
 Equal-depth (frequency) partitioning:
 It divides the range into N intervals, each containing
approximately same number of samples
 Good data scaling – good handing of skewed data
 Smoothing using Binning Methods
* Sorted data for price (in Birr): 6, 10, 11, 17, 23, 23, 26, 27,
28, 30, 31, 36
* Partition into (equi-depth) bins:
- Bin 1: 6, 10, 11, 17
- Bin 2: 23, 23, 26, 27
- Bin 3: , 28, 30, 31, 36
* Smoothing by bin means:
- Bin 1: 11, 11, 11, 11
- Bin 2: 25, 25, 25, 25
- Bin 3: 31, 31, 31, 31
* Smoothing by bin boundaries: [6,17],[23,27],[28,36]
- Bin 1: 6, 6, 6, 17
- Bin 2: 23, 23, 27, 27
- Bin 3: 28, 28, 28, 36
 Cluster Analysis
salary

cluster

outlier

age
 Data Integration
 Data integration:
 combines data from multiple sources into a coherent store
 Schema integration
 integrate metadata from different sources
 metadata: data about the data (i.e., data descriptors)
 Entity identification problem: identify real world entities
from multiple data sources, e.g., [Link]-id  [Link]-#
 possible reasons: different representations, different
scales, e.g., metric vs. British units (inches vs. cm)
 Handling Redundant
Data in Data Integration
 Redundant data occur often when integration of
multiple databases
 The same attribute may have different names in different
databases
 One attribute may be a “derived” attribute in another
table, e.g., annual revenue
 Redundant data may be able to be detected by
correlation analysis
 Careful integration of the data from multiple
sources may help reduce/avoid redundancies and
inconsistencies and improve mining speed and
quality
Correlation Analysis (Numeric Data)

 Correlation coefficient (also called Pearson’s product moment

coefficient)

i1 (ai  A)(bi  B) 

n n
(ai bi )  n AB
rA, B   i 1
(n  1) A B (n  1) A B

where n is the number of tuples, A and B are the respective

means of A and B, σA and σB are the respective standard
deviation of A and B, and Σ(aibi) is the sum of the AB cross-
product.
 If rA,B > 0, A and B are positively correlated (A’s values increase as
B’s). The higher, the stronger correlation.
 rA,B = 0: independent; rAB < 0: negatively correlated
20
 Data Transformation

 Smoothing: remove noise from data

 Aggregation: summarization, data cube
construction
 Generalization: concept hierarchy climbing
 Normalization: scaled to fall within a small,
specified range
 min-max normalization
 z-score normalization
 normalization by decimal scaling
 Attribute/feature construction
 New attributes constructed from the given ones
Normalization: Why normalization?
 Speeds-up some learning techniques (ex.
neural networks)
 Helps prevent attributes with large
ranges outweigh ones with small ranges
 Example:
 income has range 3000-200000
 age has range 10-80
 gender has domain M/F
Data Transformation: Normalization
 min-max normalization
v  minA
v'  (new _ maxA  new _ minA)  new _ minA
maxA  minA
 e.g. convert age=30 to range 0-1, when
min=10,max=80. new_age=(30-10)/(80-10)=2/7
 z-score normalization
v  meanA
v' 
stand_devA
 normalization by decimal scaling
v
v'  j Where j is the smallest integer such that Max(| v ' |)<1
10
Normalization
 Min-max normalization: to [new_minA, new_maxA]
v  minA
v'  (new _ maxA  new _ minA)  new _ minA
maxA  minA
 Ex. Let income range $12,000 to $98,000 normalized to [0.0, 1.0].
73,600  12,000
Then $73,600 is mapped to (1.0  0)  0  0.716
98,000  12,000

 Z-score normalization (μ: mean, σ: standard deviation):

v  A
v' 
 A

73,600  54,000
 Ex. Let μ = 54,000, σ = 16,000. Then  1.225
16,000
 Normalization by decimal scaling
v
v'  j Where j is the smallest integer such that Max(|ν’|) < 1
10
24
 Data Reduction Strategies
 Warehouse may store terabytes of data: Complex
data analysis/mining may take a very long time to
run on the complete data set
 Data reduction
 Obtains a reduced representation of the data set that is
much smaller in volume but yet produces the same (or
almost the same) analytical results
 Data reduction strategies
 Data cube aggregation
 Dimensionality reduction
 Data compression
 Discretization and concept hierarchy generation
 Clustering
 Partitions data set into clusters, and models it by
one representative from each cluster
 Can be very effective if data is clustered but not
if data is “smeared”
 There are many choices of clustering definitions
and clustering algorithms, further detailed in
Chapter 6
 Cluster Analysis
the distance between points in the
salary
same cluster should be small
the distance between points in different
clusters should be large
cluster

outlier

age
Discretization
 Three types of attributes:
 Nominal — values from an unordered set
 Ordinal — values from an ordered set
 Continuous — real numbers
 Discretization:
 divide the range of a continuous attribute
into intervals
 why?
 Some classification algorithms only accept
categorical attributes.
 Reduce data size by discretization
 Prepare for further analysis
Discretization and Concept hierachy
 Discretization
 reduce the number of values for a given continuous
attribute by dividing the range of the attribute into
intervals. Interval labels can then be used to replace
actual data values.
 Concept hierarchies
 reduce the data by collecting and replacing low level
concepts (such as numeric values for the attribute age)
by higher level concepts (such as young, middle-aged,
or senior).
Concept hierarchy generation for
categorical data
 Categorical attributes: finite, possibly large domain, with no
ordering among the values
 Example: Color type
 Specification of a partial ordering of attributes explicitly at
the schema level by users or experts
 Example: location is split by domain experts to
street<city<state<country
 Specification of a portion of a hierarchy by explicit data
grouping
 Specification of a set of attributes, but not of their partial
ordering
 Specification of only a partial set of attributes
Specification of a set of attributes
Concept hierarchy can be automatically
generated based on the number of distinct
values per attribute in the given attribute set.
The attribute with the most distinct values is
placed at the lowest level of the hierarchy.

country 15 distinct values

province_or_ state 65 distinct values

city 3567 distinct values

street 674,339 distinct values

 Summary
 Data quality: accuracy, completeness, consistency,
timeliness, believability, interpretability
 Data cleaning: e.g. missing/noisy values, outliers
 Data integration from multiple sources:
 Entity identification problem; Remove
redundancies; Detect inconsistencies
 Data reduction
 Dimensionality reduction; Data compression
 Data transformation and data discretization
 Normalization; Concept hierarchy generation
32
Thank You!