Mansoura University

Faculty of Computer & Information

Information System Department

Data Mining
Workbook
4th year

IS, IT, SWE, Bio

Dr. Amira Rezk

2019
Part1: Pre-Processing

 Suppose that the data for analysis includes the attribute age.
The age values are
13, 15, 16, 16, 19, 20, 20, 21, 22, 22, 25, 25, 25, 25, 30, 33, 33, 35, 35, 35, 35,
36, 40, 45, 46, 52, 70.
a) Find the mean, median, and mode of the data.
b) Give the five-number summary of the data. And Show a boxplot of the data.
c) Partition the data into three bins by each of equal-frequency and equal-
width partitioning
d) Use smoothing by bin boundaries to smooth these data
e) Use min-max normalization to transform the value 30 for age onto the range
[0:0; 1:0].
f) Use z-score normalization to transform the value 30 for age, where the SD of
age is 12.94 years.
g) Plot an equal-width histogram of width 10.

 Why do you need to perform the pre-processing operations

before perform the Data mining techniques?

 Discuss the different pre-processing operations and declare how

and why you use each one.
Part2: Mining Frequent Patterns, Associations, &
Correlations

 Suppose we have market basket data, consisting of 100

transactions and 20 items. If the support for item A is 25%, the
support for item B is 90% and the support for itemset {A, B} is
20%. Let the support and confidence thresholds be 10% and
60%, respectively.
(a) Compute the confidence of the association rule {A} → {B}. Is the rule
interesting according to the confidence measure?
(b) Compute the interest measure for the association pattern {A, B}. Describe
the nature of the relationship between item A and item B in terms of the
interest measure.
(c) What conclusions can you draw from the results of parts (a) and (b)?

 Consider the data set shown in Table 1 , Let min sup = 40%
and min conf. = 80%.
a. Find all frequent itemsets using Apriori Algorithm
b. List all the strong association rules.
c. Find the correlation the strong association rules using lift, what is the
meaning of the computed value?

Table 1: Market basket transactions

Transaction ID Items bought
1001 {i1, i4, i5}
1024 {i1, i2, i3, i5}
1012 {i1, i2, i4, i5}
1031 {i1, i3, i4, i5}
1015 {i2, i3, i5}
1022 {i2, i4, i5}
1029 {i3, i4}
1040 {i1, i2, i3}
1033 {i1, i4, i5}
1038 {i1, i2, i5}
The transaction data shown in the Table 2 from a
Table2
fast food restaurant. There are 9 distinct
Meal List of Item IDs
transactions (order:1 – order:9). There are 5 meal
Item
items that are involved in the transactions. For
Order:1 M1, M2, M5
simplicity the meal items short names (M1 – M5).
Order:2 M2,M4
The minimum support is 2/9 (.222) and the
Order:3 M2,M3
minimum confidence is 7/9 (.777).
Order:4 M1,M2.M4
a. Apply the Apriori algorithm to the dataset of Order:5 M1,M3
transactions and identify all frequent k-itemsets. Order:6 M2,M3
Show all of your work. You must show candidates Order:7 M1,M3
but can cross them off to show the ones that pass Order:8 M1,M2,M3,M5
the minimum support threshold. Note: if a Order:9 M1,M2,M3
candidate itemset is pruned because it violates the
Apriori property, you must indicate that it fails for this reason and not just
because it does not achieve the necessary support count (i.e., in these cases
there is no need to actually compute the support count). So, explicitly tag the
itemsets that are pruned due to violation of the Apriori property. (If you do not
know what the Apriori property is, do not panic. You will ultimately get the
exact same answer but will just lose a few points).

b. Find all strong association rules of the form: X  Y → Z and note their
confidence values.

Table 3:Transactions .
 Consider the data set shown in Table3 TID Items bought
1001 {i1, i4, i5}
Let min sup = 30% and min conf. = 75%. 1024 {i1, i2, i3, i5}
1012 {i1, i2, i4, i5}
a. Construct the FP-tree for these transaction 1031 {i1, i3, i4, i5}
b. Compute the support for item-sets : {i1}, 1015 {i2, i3, i5}
{i4}, {i5}, {i1;i4},{i1; i5},{i4; i5} and {i1; i4; i5} 1022 {i2, i4, i5}
1029 {i1, i3, i4}
c. Compute the confidence for the
1040 {i1, i2, i3}
association rules:
1033 {i1, i4, i5}
{i1; i4} → {i5}; {i1, i5}→ {i4} and {i4; i5} → {i1}.
1038 {i1, i2, i5}
Which one is a strong rule?
d. Compute the interest measure for the strong association rules in (c).
What is the meaning of the computed value
 Consider the data set shown in Table 4
a. Compute the support for item-sets
{i5}, {i2; i4}, and {i2; i4; i5} by treating each transaction ID as a market basket.
b. Use the results in (a) to compute the confidence for the association rules
{i2; i4} → {i5} and {i5} → {i2; i4}.
c. Is confidence a symmetric measure?
d. Repeat part (a) by treating each customer ID as a market basket. Each
item should be treated as a binary variable (1 if an item appears in at least one
transaction bought by the customer, and 0 otherwise.)
e. Use the results in part (d) to compute the confidence for the association
rules {i2; i4} → {i5} and {i5} → {i2; i4}.
f. Discuss whether there are any relationships between support and
confidence of parts {a, b} and {d, e}.
g. Compute the lift for the association rules {i2; i4} → {i5} and {i5} → {i2; i4}
in parts {b, e}. what is the meaning of the computed value.

Table4: Market basket transactions.

Customer Transaction Items
ID ID bought
1 1001 {i1, i4, i5}
1 1024 {i1, i2, i3, i5}
2 1012 {i1, i2, i4, i5}
2 1031 {i1, i3, i4, i5}
3 1015 {i2, i3, i5}
3 1022 {i2, i4, i5}
4 1029 {i3, i4}
4 1040 {i1, i2, i3}
5 1033 {i1, i4, i5}
5 1038 {i1, i2, i5}
 The following table summarizes supermarket transaction data, where hot
dogs refers to the transactions containing hot dogs, hot dogs refers to
the transactions that do not contain hot dogs, hamburgers refers to the
transactions containing hamburgers, and hamburgers refers to the
transactions that do not contain hamburgers.

a. Suppose that the association rule “hot dogs → hamburgers” is mined.

Given a minimum support threshold of 25% and a minimum confidence
threshold of 50%, is this association rule strong?

b. Based on the given data, is the purchase of hot dogs independent of the
purchase of hamburgers? If not, what kind of correlation relationship exists
between the two? Note: the value needed to reject the hypothesis is 10.828

 A database has four transactions. Let min sup = 60% and min
conf = 80%.
Part 3: Classification
Table1
 Consider the sample data shown in Instance A1 A2 A3 class

Table1,for a binary classification 1 T T 1.0 +

2 T T 6.0 +
problem. 3 T F 5.0 -
(a) What is the entropy of this collection of data 4 F F 4.0 +
with respect to the positive class? 5 F T 7.0 -
(b) What are the information gains of a1 and a2 6 F T 3.0 -
relative to this data? 7 F F 8.0 -
(c) What is the best split (among a1 and a2) 8 T F 7.0 +
according to the information gain? 9 F T 5.0 -

 Consider the following decision tree

a. Extract IF-THEN rules from this decision tree
b. Evaluate this classification model using a test data set in table2;
Calculate the following measurements: accuracy, error rate, sensitivity,
specificity, and precision

Table2: test data set

Class:
RID age income student Credit_rating
buys_computer
1 youth high no fair yes
2 youth high no excellent no
3 youth medium no fair no
4 middle aged high no fair yes
5 middle aged low yes excellent yes
6 middle aged high yes fair no
7 senior low yes excellent no
8 senior low yes fair yes
9 senior medium no fair no
10 senior medium yes fair yes
 Table3 presents a training set D of class-labeled tuples
randomly selected from customer database. The two decision
classes are buy from shop (YES) or does not buy from shop
(NO).
a. Generate a decision tree from the training tuples of data partition, D
b. Extract IF_THEN rules from a decision tree. How to assess the goodness
of a rule?
c. Use Naïve Bayesian Classification method to classify an individual who
has the following attributes: with class Yes =(4/4) * (0/4) * (1/4) * probability of class
and repeat for no then take the big
{Car ownership=”N0”; Marital status= “married”, taxable income= medium}
Table3: data set
Taxable Car Marital Buy from
income ownership status shop
high Yes Single No
medium No Married No
low No Single Yes
medium Yes Married No
medium No Divorced Yes
low No Married No
high Yes Divorced No
low No Single Yes
low No Married No
low No Single Yes

Table4
 Consider the sample data shown in Table4 A1 A2 A3 class
T T P +
a- What is the entropy of this collection of data T T P +
b- What are the information gains of A1, A2 and A3 T F N -
F F N +
relative to this data? F T N -
F T N -
c- What is the best split according to the information F F P -
T F P +
gain in part (b)? F T N -
d- Use Naïve Bayesian Classification method to classify T F N -
an object which has the following attributes: {A1:T, A2:F,
A3:N}
 For a given data in table5, count represents the number of data
tuples having the values for department, status, age, and salary.
Let status be the class label attribute.
a- What is the entropy of this collection of data .
b- What are the information gains of department, age, and salary
c- What is the best split according to the information gain in (b)?
d- Use Naïve Bayesian Classification method to classify an individual who has
the following attributes:
{ department =”marketing”; age= “youth”, salary= low}

Table5
department age salary status count
sales Middle aged medium senior 30
sales youth low junior 30
sales Middle aged low junior 40
systems youth medium junior 20
systems Middle aged high senior 20
systems senior high senior 10
marketing senior medium senior 10
marketing Middle aged medium junior 20
secretary senior medium senior 10
secretary youth low junior 10
Part4 : Clustering

 Suppose that the data mining task is to cluster the following

nine points (with (x, y) representing location) into three clusters,
the distance function is Euclidean distance. Suppose initially we
assign A1, B1, and C1 as the center of each cluster.

A1(2, 10), A2(2, 5), A3(8, 4), B1(5, 8), B2(7, 5), B3(6, 4), C1(1,
2), C2(4, 9), C3(2,8):

- Use the k-means algorithm to show the three cluster centers after the
first-round execution

 For the following points

(2, 5), (3, 10), (8, 4), (5, 8), (8, 5), (6, 4), (1, 2), (4, 9):
Assume that k = 2 and initially assign (2, 5) and (3, 10) as the
center of each cluster.
- Apply the k-means algorithm until the clusters do not change, using the
Manhattan distance.
(Hint: The Manhattan distance is: d(i, j) = |xi1-xj1|+ |xi2-xj2|+ ….+ | xin-xjn|.)

 Suppose that the data mining task is to cluster the following

points (with (x, y) representing location), The distance function
is the Manhattan distance. Suppose initially we assign A1, B1,
and C1 as the center of each cluster.
A1(1, 2), A2(7, 5), A3(6, 4), A4 (3, 4),
B1 (5, 8), B2(2, 5), B3(8, 4),
C1(2, 10), C2(4, 9), C3(2,8)
- Use the k-means algorithm to show the three cluster centers after the
first round execution. (Hint: The Manhattan distance is: d(i, j) = |xi1-xj1|+
|xi2-xj2|+ ….+ | xin-xjn|.)
 For the following nine points (1,2), (2,2), (2,3), (3,4), (4,3), (5,4),
(6,6), (7,5), (8,4). Assume that k = 2 and initially assign (2,2)
and (5,4) as the center of each cluster.
- Apply the k-means algorithm using the Manhattan distance and show the
new cluster centers after the first round execution (Hint: The Manhattan
distance is: d(i, j) = |xi1-xj1|+ |xi2-xj2|+ ….+ | xin-xjn|.)
- Compute the silhouette coefficient for object (3, 4). What is the meaning
of the computed value?

 For the following nine points (2,2), (2,7), (3,1), (3,5), (4,3), (4,8),
(5,2), (6,2), (6,5). Assume that k = 2 and initially assign (2,2)
and (2,7) as the center of each cluster.
- Apply the k-means algorithm using the Manhattan distance and show the
new cluster centers after the first-round execution
- Suppose that your result in (a) is the final cluster, how can you use it to
detect the outlier? what is the object which likely be an outlier?

 Describe the principles and ideas regarding Agglomorative

Hierarchical Clustering. - Show the different steps of the
algorithm using the dissimilarity matrix below and complete link
clustering. Give partial results after each step.
1 2 3 4 5
1 0
2 2 0
3 4 3 0
4 10 7 9 0
5 8 5 6 1 0
Choose the Correct Answer.
1. For the following association rule: Computer → Webcam (60%, 100%):
Which of the following is true?
I. 100% of costumers bought both a computer and a webcam
II. 60% of costumers bought both a computer and a webcam
III. 100% of costumers who bought a computer bought also a webcam
IV. 60% of costumers who bought a computer bought also a webcam
a. II only b III Only c. I and IV d. II and III
2. We have Market Basket data for 1,000 rental transactions at a Video Store.
There are four videos for rent -- Video A, Video B, Video C and Video D. The
probability that both Video C and Video D are rented at the same time is known
as ________ .
a. Correlation b. support c. lift d. confidence

Consider the following transaction database: Suppose that minsup is set to

40% and minconf. to 70%.
TransID Items
3. The support of the item set A, B, E is……… T100 A, B, C, D
T200 A, B, C, E
a. 50% b.40% c. 70% d. 66%
T300 A, B, E, F, H
4. Based on the given minimum support the item set T400 A, C, H
A,B,E is……..
a. frequent b. not frequent c. strong d. not strong
5. The confidence of the rule A, B → E is 2/3
a. 50% b.40% c. 100% d. 66%
6. Based on the given minimum confidence the rule A, B → E is……..
a. frequent b. not frequent c. strong d. not strong
7. The lift of the rule A, B → E is……. 0.5/(0.75 * 0.5) =
a. 1.33 b.1 c. 0.89 d. 0.66
8. The value of the lift in the previous question means that items are……….
a. positive correlated b. negative correlated
c. independent d. strong
9. For the given data {33, 25, 42, 25, 31, 37, 46, 29, 38} the five numbers
summery will be … 25 27 33 40 46
a. 25, 27, 32, 35, 46 b. 25, 27, 33, 35, 46 c. 14, 27, 33, 35, d.
19, 29, 32, 38, 43
10. If you use min-max normalization to transform the value 33 onto the range
[1.0, 2.0] the new value is
a. 0.38 b. 1.38 c. 0.038 d. 1.038
11. Identify the outlier for the given data? 23, 34, 27, 7, 30, 26, 28, 31, 34
a. 7 b. 23 c. 31 d. 34
From the given Confusion Matrix
12. Accuracy is……. (9654 + 2588)/10000 Confusion Matrix
a.0.99 b.0.95 c.0.86 d.0.05 Predicted Total
13. Error rate is……. Yes No
a.0.99 b.0.95 c.0.86 d.0.05
Yes 6954 46 7000
TP FN
14. Sensitivity is……. TP/P 6954/7000

Actual
No 412 2588 3000
a.0.99 b.0.95 c.0.86 d.0.05 FP TN
Total 7366 2634 10000
15. Specificity is ……….
TN/N = 2588/3000
a.0.99 b.0.95 c.0.86 d.0.05
Assume, you want to cluster 7 observations into 3 clusters using K-Means
clustering algorithm. After first iteration clusters, C1, C2, C3 has following
observations: C1: {(2, 2), (4, 4), (6, 6)} C2: {(0, 4), (4, 0)} C3: {(5, 5), (9, 9)}
16. What will be the cluster centroids if you want to proceed for second
iteration?
a. C1: (4, 4), C2: (2, 2), C3: (7, 7) b. C1: (6, 6), C2: (4, 4), C3: (9, 9)
c. C1: (2, 2), C2: (0, 0), C3: (5, 5) d. None of these
17. What will be the Manhattan distance for observation (9, 9) from cluster
centroid C1 in second iteration? (9-4) + (9-4)

a. 10 b. 5 c. 13 d. None of these
18. Consider the given data: {3, 4, 5, 10, 21, 32, 43, 44, 46, 52, 59, 67}, Using
equal-width partitioning and four bins, how many values are there in the first
bin?
a. 3 b. 4 c. 5 d. 6
19. If smooth by median is applied to the previous bins, what is the new value
of the data in the first bin?
a. 4 b. 4.5 c. 5 d. 7.5
20. Which of the following lists all parts of the five-number summary?
a. Mean, Median, Mode, Range, and Total
b. Minimum, Quartile1, Median, Quartile3, and Maximum
c. Smallest, Q1, Q2, Q3, and Q4
d. Minimum, Maximum, Range, Mean, and Median
Answer the following Questions
1. Define: A centroid in k-means.
2. Define: A core point in DBSCAN
3. Define: association and correlation analysis. Give an example
4. Define: cluster analysis. Give an example
5. Define: Data Cleaning, Data integration, Data reduction, Data
transformation, Discretization
6. Define: outlier analysis. Give an example
7. Define: regression. Give an example
8. Give an example for nonparametric data reduction strategies.
9. Give an example for parametric data reduction strategies.
10. How does K-means differ from DBSCAN
11. How to assess the goodness of a rule?
12. How you can solve missing values problems
13. If a person’s height is measured in inches then what kind of attribute you
will use? RATIONAL
14. If the correlation coefficient of the items bred and rice is equal to 1.5. This
means what?
15. If the covariance of the items bred and rice is equal to 1. This means
what? positive
16. If the information gain of age and income attributes are 0.24 and 0.024
respectively which one you will chose as the splitting attribute
17. If the lift measure of the items bred and rice is equal to 0.5. This means
what?
18. If the lift measure of the items bred and rice is equal to 1. This means
what?
19. If the lift measure of the items bred and rice is equal to 1.5. This means
what?
20. If the mean is equal to the median then this might be an indication that
the data is what?
21. If the mean is larger than the median then this might be an indication that
the data is what?
22. If the mean is smaller than the median then this might be an indication
that the data is what? negative
23. If you have 100 values in my data and I add 5.0 to all of the values, then
how will this change the median? value is increased ...but position still the same

24. If you have 100 values in my data and I add 5.0 to all of the values, then
how will this change the median?
25. List the Cluster Analysis Methods
26. List the Major Preprocessing Tasks That Improve Quality of Data
27. List the steps of knowledge discovery
28. List the transformation strategies binning, regression

29. List the types of outliers. Give an example for each one.
30. The confidence for the association rule {bread} → {milk, diapers} was
determined to be 0.95. What does the value 0.95 mean?
31. The support for the association rule {bread} → {milk, diapers} was
determined to be 0.95. What does the value 0.95 mean?
32. What are rules conflicts? How can you solve it?
33. What are the data smoothing techniques? binning, regression, outlier analysis

34. What are the different strategies of data reduction?

35. What are the main advantages and disadvantages of Decision Tree
classification algorithms?
36. What are the terminating conditions in decision tree induction?
37. What classifiers are normally considered to be easy to interpret?
38. What clustering algorithms can find clusters of arbitrary shape?
39. What data mining task should be used to detect fraudulent usage of credit
cards?
40. What is over fitting? Briefly describe one method to prevent over-fitting in
classification trees.
41. What is the Apriori property?
42. What is the bootstrap sampling?
43. What is the different between noise and outlier?
44. What is the different between symmetric and asymmetric binary
attribute?
45. What is the five numbers summery of the data? How is it represented
graphically?
46. What is the majority voting? When you use it?
47. What is the means of association rule computer → webcam (60%, 100%)
48. What is the mode of the data? What is the mean of (bimodal, trimodel)
49. What is the problem that related to calculate the mean? How you can fix
it?
50. What is the problem that related to use global constant to fill in the
missing values.
51. What is the redundant attribute? How can you detect it?
52. What Kinds of Data Can Be Mined?
53. What Kinds of Patterns Can Be Mined?
54. When are objects q & m density-connected.
55. When is object p density-reachable from another object q?
True or False
56. The silhouette coefficient is a method to determine the natural number of
clusters for hierarchical algorithms density-based algorithms F evaluate of the cluster
57. All continuous variables are ratio F ratio and interval

58. Association rules provide information in the form of "if-then" statements.

F classification is if then statement
59. Attributes are sometimes called variables and objects are sometimes
called observations T
60. Binary variables are sometimes continuous F

61. Cluster is the process of finding a model that describes and distinguishes
data classes or concepts. F cluster --> classification
62. Computing the total sales of a company. Is a data mining task? F

63. Correlation analysis divides data into groups that are meaningful, useful,
or both. F ..correlation analysis --> cluster analysis
64. Database mining refers to the process of deriving high-quality information
from text. F ..database mining --> text mining
65. Dissimilarity matrix stores n data objects that have p attributes as an n-by-
p matrix F .. data matrix
66. Dividing the customers of a company according to their profitability. is a
data mining task? F
67. For an association rule, if we move one item from the right-hand-side to F because
confidence is not
the left-hand-side of the rule, then the confidence will never change. symmetric

68. If all the proper subsets of an itemset are frequent, then the itemset itself
must also be frequent. F
69. In decision tree algorithms, attribute selection measures are used to rank
attributes T
70. In decision tree algorithms, attribute selection measures are used to
reduce the dimensionality T
71. In lazy learner we interest in the largest distance. F the smallest distance

72. Intrinsic methods measure how well the clusters are separated T
73. Multimedia Mining is the application of data mining techniques to
discover patterns from the Web. F multimedia mining --> web mining
74. Regression is a method of integration F except integration
75. Strategies for data transformation include chi-square test F chi-square is in integration

76. The Pruning make the decision tree more complex F complex --> simple
77. An object is an outlier if its density is equal to the density of its neighbors. F equal to -->
less than
78. A common weakness of association rule mining is that it is not produce
enough interesting rules F
79. Accuracy is interestingness measures for association rules F Accuracy --> support and
confidence
80. Binning is a method of reduction F (binning is in Transformation $ cleaning)

81. Core object is an object whose -neighborhood contains objects less

than MinPts less than --> more than

82. Correlation analysis is used to eliminate misleading rules. T

83. Correlation is a method of cleaning F correlation in Transformation

84. Data matrix stores a collection of proximities for all pairs of n objects as an
n-by-n matrix F Data matrix --> dissimilarity matrix
85. Extracting the frequencies of a sound wave. Is a data mining task? F

86. Incomplete data problem can be solved by binning F incomplete data --> noise data

87. K-Nearest Neighbor Classifiers do classification when new test data is

available T
88. Median is a value that occurs most frequently in the attribute values F mean --> mode

89. Mode is a middle value in set of ordered values F mode --> median

90. One strength of a Bayesian Classifier is that it can be easily trained T

91. Outlier analysis is a method of transformation transformation --> cleaning

92. Predicting the outcomes of tossing a (fair) pair of dice. Is a data mining
task? F
93. Recall is interestingness measures for association rules F Recall --> lift

94. Redundancy is an important issue in data cleaning F data cleaning --> data integration

95. Sampling methods smooth noisy data F sampling --> regression or outlier analysis or binning

96. Sorting a student database based on student identification numbers. Is a

data mining task? F

97. The bottleneck of the Apriori algorithm is caused by the number of

association rules F association rules --> number of candidates
98. The goal of clustering analysis is to maximize the number of clusters F

99. the object is local outlier if it is deviate significantly from the rest of the
dataset F local --> global outlier
100. The silhouette coefficient is a method to determine the natural number of
clusters for hierarchical algorithms F determine the evaluation of the cluster not determine the
number of clusters

My best wishes;