Data Mining:

Concepts and Techniques

— Chapter 2 —

Slides Curtesy of Textbook

1
Chapter 2: Getting to Know Your Data

Data Objects and Attribute Types

Basic Statistical Descriptions of Data

Data Visualization

Measuring Data Similarity and Dissimilarity

Summary

2
 Types of Data Sets
Record
Relational records
Data matrix, e.g., numerical matrix,
crosstabs
Document data: text documents: term-
frequency vector
Transaction data
Graph and network
World Wide Web
Social or information networks
Molecular Structures
Ordered TID Items
Video data: sequence of images 1 Bread, Coke, Milk
Temporal data: time-series
2 Beer, Bread
Sequential Data: transaction sequences
3 Beer, Coke, Diaper, Milk
Genetic sequence data
Spatial, image and multimedia: 4 Beer, Bread, Diaper, Milk
Spatial data: maps 5 Coke, Diaper, Milk
Image data:
Video data:
3
Important Characteristics of Structured Data

Dimensionality
Curse of dimensionality
Sparsity
Only presence counts
Resolution
Patterns depend on the scale
Distribution
Centrality and dispersion

4
 Data Objects

Data sets are made up of data objects.
A data object represents an entity.
Examples: 
sales database:  customers, store items, sales
medical database: patients, treatments
university database: students, professors, courses
Also called samples , examples, instances, data points, objects, 
tuples.
Data objects are described by attributes.
Database rows ‐> data objects; columns ‐>attributes.
5
 Attributes
Attribute (or dimensions, features, variables): a data 
field, representing a characteristic or feature of a data 
object.
E.g., customer _ID, name, address
Types:
Nominal
Binary
Ordinal
Numeric: quantitative
Interval‐scaled
Ratio‐scaled 6
 Attribute Types
Nominal: categories, states, or “names of things”
Hair_color = {auburn, black, blond, brown, grey, red, white}
marital status, occupation, ID numbers, zip codes
Binary
Nominal attribute with only 2 states (0 and 1)
Symmetric binary: both outcomes equally important
e.g., gender
Asymmetric binary: outcomes not equally important.  
e.g., medical test (positive vs. negative)
Convention: assign 1 to most important outcome (e.g., HIV 
positive)
Ordinal
Values have a meaningful order (ranking) but magnitude between 
successive values is not known.
Size = {small, medium, large}, grades, army rankings

7
 Numeric Attribute Types
Quantity (integer or real‐valued)
Interval
Measured on a scale of equal‐sized units
Values have order
E.g., temperature in C˚or F˚, calendar dates
No true zero‐point
Ratio
Inherent zero‐point
We can speak of values as being an order of magnitude 
larger than the unit of measurement (10 K˚ is twice as 
high as 5 K˚).
e.g., temperature in Kelvin, length, counts, 
monetary quantities
8
 Discrete vs. Continuous Attributes
Discrete Attribute
Has only a finite or countably infinite set of values
E.g., zip codes, profession, or the set of words in a 
collection of documents 
Sometimes, represented as integer variables
Note: Binary attributes are a special case of discrete 
attributes 
Continuous Attribute
Has real numbers as attribute values
E.g., temperature, height, or weight
Practically, real values can only be measured and 
represented using a finite number of digits
Continuous attributes are typically represented as floating‐
point variables
9
Chapter 2: Getting to Know Your Data

Data Objects and Attribute Types

Basic Statistical Descriptions of Data

Data Visualization

Measuring Data Similarity and Dissimilarity

Summary

10
Basic Statistical Descriptions of Data
Motivation
To better understand the data.
Central tendency– the center, the representative.
Dispersion– variation, spread.

11
 Measuring the Central Tendency
n
Mean (algebraic measure) (sample vs. population): 1 x
x xi
Note: n is sample size and N is population size.  n i 1 N
n
Weighted arithmetic mean:
wi xi
Trimmed mean: chopping extreme values x i 1
n
Median:  wi
i 1
Middle value if odd number of values, or average of the 
middle two values otherwise
Estimated by interpolation (for grouped data):
n/2 ( freq ) l Median
median L1 ( ) width interval

Mode freqmedian
Value that occurs most frequently in the data
Unimodal, bimodal, trimodal
Empirical formula: mean mode 3 (mean median)
12
Symmetric vs. Skewed Data
Mean
Median
Mode
Median, mean and mode of  symmetric
symmetric, positively and negatively 
skewed data

positively skewed negatively skewed

August 27, 2015 Data Mining: Concepts and Techniques 13

 Measuring the Dispersion of Data
Quartiles, outliers and boxplots
Quartiles: Q1 (25th percentile), Q3 (75th percentile)
Inter‐quartile range: IQR = Q3 – Q1 
Five number summary: min, Q1, median, Q3, max
Boxplot: ends of the box are the quartiles; median is marked; add whiskers, 
and plot outliers individually
Outlier: usually, a value higher/lower than 1.5 x IQR
Variance and standard deviation (sample: s, population: σ)
Variance: (algebraic, scalable computation)
n n
2 1 2 1 2 1 n 2 1 n
1 n
s ( xi x) [ xi ( xi ) ] 2
( xi ) 2
xi
2 2
n 1i 1 n 1 i1 n i1 N i 1 N i 1

Standard deviation s (or σ) is the square root of variance s2 (or σ2)

14
 Boxplot Analysis

Five‐number summary of a distribution
Minimum, Q1, Median, Q3, Maximum
Boxplot
Data is represented with a box
The ends of the box are at the first and third 
quartiles, i.e., the height of the box is IQR
The median is marked by a line within the box
Whiskers: two lines outside the box extended 
to Minimum and Maximum
Outliers: points beyond a specified outlier 
threshold, plotted individually

15
 Visualization of Data Dispersion: 3-D Boxplots

August 27, 2015 Data Mining: Concepts and Techniques 16

 Properties of Normal Distribution Curve

The normal (distribution) curve
From μ–σ to μ+σ: contains about 68% of the measurements  
(μ: mean, σ: standard deviation)
From μ–2σ to μ+2σ: contains about 95% of it
From μ–3σ to μ+3σ: contains about 99.7% of it

68% 95% 99.7%

−3 −2 −1 0 +1 +2 +3 −3 −2 −1 0 +1 +2 +3 −3 −2 −1 0 +1 +2 +3

17
Graphic Displays of Basic Statistical Descriptions

Boxplot: graphic display of five‐number summary
Histogram: x‐axis are values, y‐axis repres. frequencies 
Quantile plot:  each value xi is paired with fi  indicating that 
approximately 100 fi % of data  are  xi
Quantile‐quantile (q‐q) plot: graphs the quantiles of one 
univariant distribution against the corresponding quantiles of 
another
Scatter plot: each pair of values is a pair of coordinates and 
plotted as points in the plane

18