Slides for the 2016/2017 edition of the Data Mining and Text Mining Course at the Politecnico di Milano. The course is also part of the joint program with the University of Illinois at Chicago.
1. Prof. Pier Luca Lanzi
Data Exploration
Data Mining andText Mining (UIC 583 @ Politecnico di Milano)
2. Prof. Pier Luca Lanzi
Readings
• “Data Mining and Analysis” – Chapter 2 & 3
• Loading Data and Basic Formatting in R
http://flowingdata.com/2015/02/18/loading-data-and-basic-formatting-in-r/
2
3. Prof. Pier Luca Lanzi
http://www.kdnuggets.com/2015/03/cartoon-us-chief-data-scientist-challenge.html
4. Prof. Pier Luca Lanzi
Before trying anything,
you should know your data!
5. Prof. Pier Luca Lanzi
What is Data Exploration?
• It is the preliminary exploration of the data aimed at identifying
their most relevant characteristics
• What the key motivations?
§Helping to select the right tool for preprocessing and data
mining
§Exploiting humans’ abilities to recognize patterns
not captured by automatic tools
• Related to Exploratory Data Analysis (EDA), created by
statistician John Tukey
5
6. Prof. Pier Luca Lanzi
Exploratory Data Analysis
• “An approach of analyzing data to summarize their main
characteristics without using a statistical model or having
formulated a prior hypothesis.”
• “Exploratory data analysis was promoted by John Tukey
to encourage statisticians to explore the data, and
possibly formulate hypotheses that could lead to new
data collection and experiments.”
• Seminal book is “Exploratory Data Analysis” by Tukey
• Nice online introduction in Chapter 1 of the NIST
Engineering Statistics Handbook
http://www.itl.nist.gov/div898/handbook/index.htm
• http://en.wikipedia.org/wiki/Exploratory_data_analysis
6
7. Prof. Pier Luca Lanzi
http://vis.stanford.edu/files/2012-EnterpriseAnalysisInterviews-VAST.pdf
8. Prof. Pier Luca Lanzi
What Data Exploration Techniques?
• Exploratory Data Analysis (as originally defined by Tukey),
was mainly focused on
§Visualization
§Clustering and anomaly detection
(viewed as exploratory techniques)
§In data mining, clustering and anomaly detection are major
areas of interest, and not thought of as just exploratory
• In this section, we focus on data exploration using
§Summary statistics
§Visualization
• We will come back later to data exploration when discussing
clustering
8
19. Prof. Pier Luca Lanzi
Geometric Interpretation of
Correlation
• The correlation coefficient is simply the cosine of the angle
between the two centered attribute vectors
19
23. Prof. Pier Luca Lanzi
Summary Statistics
• They are numbers that summarize properties of the data
• Summarized properties include location, mean, spread, skewness,
standard deviation
• Most summary statistics can be calculated in a single pass
23
24. Prof. Pier Luca Lanzi
Frequency and Mode
• The frequency of an attribute value is the percentage of time the
value occurs in the data set
• For example, given the attribute ‘gender’ and a representative
population of people, the gender ‘female’ occurs about 50% of
the time.
• The mode of an attribute is the most frequent attribute value
• The notions of frequency and mode are typically used with
categorical data
24
25. Prof. Pier Luca Lanzi
Measures of Location:
Mean and Median
• The mean is the most common measure of the location of a set
of points
• However, the mean is very sensitive to outliers
• Thus, the median or a trimmed mean is also commonly used
25
26. Prof. Pier Luca Lanzi
Measures of Spread:
Range and Variance
• Range is the difference between the max and min
• The variance or standard deviation is the most common measure
of the spread of a set of points
• However, this is also sensitive to outliers, so that other measures
are often used
26
27. Prof. Pier Luca Lanzi
Percentiles
• For continuous data, the notion of a percentile is veryuseful
• Given an ordinal or continuous attribute x and a number p
between 0 and 100, the p-th percentile is a value xp of x such
that p% of the observed values of x are less than xp
• For instance, the 50th percentile is the value x50% such that 50%
of all values of x are less than x50%
27
28. Prof. Pier Luca Lanzi
Visualization
• Visualization is the conversion of data into a visual or tabular
format so that the characteristics of the data and the relationships
among data items or attributes can be analyzed or reported
• Data visualization is one of the most powerful and appealing
techniques for data exploration
§Humans have a well developed ability to analyze large
amounts of information that is presented visually
§Can detect general patterns and trends
§Can detect outliers and unusual patterns
28
29. Prof. Pier Luca Lanzi
Barplots
• They use horizontal or vertical bars to compare categories.
• One axis shows the compared categories, the other axis
represents a discrete value
• Some bar graphs present bars clustered in groups of more than
one (grouped bar graphs), and others show the bars divided into
subparts to show cumulate effect (stacked bar graphs)
29
30. Prof. Pier Luca Lanzi
Histograms
• They are a graphical representation of the distribution of data
introduced by Karl Pearson in 1895
• They estimate the probability distribution of a continuous variables
• They are representations of tabulated frequencies depicted as adjacent
rectangles, erected over discrete intervals (bins). Their areas are proportional
to the frequency of the observations in the interval.
30
31. Prof. Pier Luca Lanzi
Histograms
• The height of each bar indicates the number of objects
• Shape of histogram depends on the number of bins
31
Example: Petal Width (10 and 20 bins, respectively)
32. Prof. Pier Luca Lanzi
Maps
• Maps are a very effective communications tool that can visualize
thousands of data points in just one figure
32
33. Prof. Pier Luca Lanzi
Mapping Data for Visualization
• Data objects, their attributes, and the relationships among data
objects are translated into graphical elements such as points, lines,
shapes, and colors
• Objects are often represented as points
• Their attribute values can be represented as the position of the
points or the characteristics of the points, e.g., color, size, and
shape
• If position is used, then the relationships of points, i.e., whether
they form groups or a point is an outlier, is easily perceived.
33
34. Prof. Pier Luca Lanzi
Data Arrangement
• Is the placement of visual elements within a display
• Influence on how easy it is to understand the data
34
35. Prof. Pier Luca Lanzi
Selection
• Is the elimination or the de-emphasis of
certain objects and attributes
• Selection may involve choosing a subset of attributes
§Dimensionality reduction is often used to reduce the number
of dimensions to two or three
§Alternatively, pairs of attributes can be considered
• Selection may also involve choosing a subset of objects
§ A region of the screen can only show so many points
§Can sample, but want to preserve points in sparse areas
35
36. Prof. Pier Luca Lanzi
Box Plots
• Box Plots (invented by Tukey)
• Another way of displaying the distribution of data
• Following figure shows
the basic part of a box plot
• IQR is the interquartile range
or Q3-Q1
36
outlier
Min (> Q1 -1.5 IQR)
1st quartile (Q1)
3rd quartile (Q3)
2nd quartile (Q2)
Max ( < Q3 +1.5 IQR)
38. Prof. Pier Luca Lanzi
Scatter Plots
• Attributes values determine the position
• Two-dimensional scatter plots most common, but can have
three-dimensional scatter plots
• Often additional attributes can be displayed by using the size,
shape, and color of the markers that represent the objects
• It is useful to have arrays of scatter plots can compactly
summarize the relationships of several pairs of attributes
• Examples: http://www.statmethods.net/graphs/scatterplot.html
38
42. Prof. Pier Luca Lanzi
Scatter Plot Combined with Box Plots 42
43. Prof. Pier Luca Lanzi
Contour Plots
• Useful for continuous attributes measured on a spatial grid
• They partition the plane into regions of similar values
• The contour lines that form the boundaries of these regions
connect points with equal values
• The most common example is contour maps of elevation
• Can also display temperature, rainfall, air pressure, etc.
43
45. Prof. Pier Luca Lanzi
Matrix Plots
• Can plot the data matrix
• This can be useful when objects are sorted according to class
• Typically, the attributes are normalized to prevent one attribute
from dominating the plot
• Plots of similarity or distance matrices can also be useful for
visualizing the relationships between objects
45
47. Prof. Pier Luca Lanzi
Other Visualization Techniques
• Star Plots
§Similar approach to parallel coordinates,
but axes radiate from a central point
§The line connecting the values of an object is a polygon
• Chernoff Faces
§Approach created by Herman Chernoff
§This approach associates each attribute with a characteristic of
a face
§The values of each attribute determine the appearance of the
corresponding facial characteristic
§Each object becomes a separate face
§Relies on human’s ability to distinguish faces
47
54. Prof. Pier Luca Lanzi
do we read the articles we share?
does my banner work?
55. Prof. Pier Luca Lanzi
“When we combined attention and traffic to find the story that had
the largest volume of total engaged time, we found that it had fewer
than 100 likes and fewer than 50 tweets. Conversely, the story with
the largest number of tweets got about 20% of the total engaged
time that the most engaging story received.”
56. Prof. Pier Luca Lanzi
“Here’s the skinny, 66% of attention on a normal media page is spent
below the fold.That leaderboard at the top of the page? People
scroll right past that and spend their time where the content not the
cruft is. .”
57. Prof. Pier Luca Lanzi
http://time.com/12933/what-you-think-you-know-about-the-web-is-wrong/?iid=biz-category-mostpop2
58. Prof. Pier Luca Lanzi
http://apps.startribune.com/news/20140313-beer-me-minnesota/