Data Mining

1 Homework+ 1 midterm =
-
% 60

final Exam =- 10 %

KDD
O LAB
D ala Warehouse
Data Cube

Object Relational Databases

Temporal Database
Data base
Sequence
time series Database

Data Mining

X
DescriptivePredictive

must be now
Chapter z : cas function less
 variances
standard deviation and

= (x-X

Wavele
Transformation
2 .
5:4 Numerosity Reduction

t
 Name : Mohamed Osman Hassan
Student Number: 210 S13737

department : Software Engineering

Signature :
At
Question L answer : We can
find mean by adding all

data and dividing in with the number data

of
=

= E Xi
83 31.
= =

14 18 19 00 20
, ,
16 , 16
, , , ,
21 , 22 , 22 23
, ,
25, 25 25 30,
, , 33 , 33, 35, 35
, ,35, 36, 40, 45, 46, 52
35
,
70 , 85

Median 27 5
230
= = .

Question 10 : The
highest frequency of data set

and only
there
mode.
is 35
modality is unimodal Win one

Question 1c :
Midrange ishighest
value + lowest valued
2

+ 149 5 .

Question 1D
firs quartile (o) is
14 18 16 , 16 19 00 20
, 21 , 22 , 22 23 25, 25 as
, , , , , , , ,

average =

(2) = 20 .

5
Q3 is : 30, 33
, 33, 35, 35, 35;35, 36, 40 45
, ,
46, 52 ,
70 , 85

Q3
average
(336) 35:5
= = -

Question -E ; it menus we have to write Min

, R , median and=
= 14 20 . 5 27 5 35 5
.
.
, , ,

Question If :

: 18
E

Question 19 : Grantile plac Ordersalata from smallest to

largest .

Quantile-Qantile Plot : Calculates the quarties of

dow data sets

One theoretical one dataset .

,
 Question 2 a : cosine similarity
ddek
equation
dini = = 0 .
0660 = co
similaritie

correlation

equation
,
*
=
8 375
j
:
2+ 2 = 1 5
20
+
.
=

2 (xi - *
) (yij) (0) (0 5) + 0 (-1 5) + 0
=
. . :
.

00 5) + . 0. 10 .

5) = 0

0 .

[10 2s)
. + 12 253 + (0 25) + 20
. .
. 25)]
Of undefined
=
=
correlation

-
Sodedian distance : Jaxi-yi)2 =
J(1-2 + 11-0+ 11-2)+ (1-252 = 2

02(b)
= did = =
0 . 516 = Cosine
Similarity

T = 0 . 75
2 (xi - <
) (y ,j) = (0 - 0. 75)(1-0 75) + 21 -0 75710
.
. -
0 .

75) +
y = 0 . 75

(0 -0 75) (1-0 75) . .

+ 12 -
0 .
75)(1-0 75) :

=
- 0 .
1875 -0 1875 -
-
0. 1875 + 0 3175 .

= -
0 75
75)"(0 29)
.

(xi <y [ (yi j7 7592(8 25)2 + (0 2572( 07512 + 1- 0

.

= ( 0
-
- - :
- . . .

↑ (102512 (0 25)2 .

195
-0
.

C-0110-11Tar
Zocedian distance :
J + = 2

Jaccard Similarity :y ==

Ge() y
=

t
=
0408

=
zocedian distance
F : Th to = Je = 1 752
.

Jaccard Similarity
:
 God
No
Cosine similarity
==
0

T
= = 0 ,
83 = 0
a

=
correlation
0 343

&=
.

Jaccard Similarity
: 2
= 0 .
667

Q2e

No
Cosine similarity =
=
correlation 0

T = 0

Y = 2

Jaccard Similarity
:= 0

Py(a) Evclidean distance = (18-201+ (213+ (42 37) + 16-4) -

=
n + 4 + 25 + y 557 6 08 = = .

Manhattan distance 118-201

= + (21 + 142 371 + 16-41
-

= 11

Minhowski distance = 18-20 23

+ 42-37'+ 6-13
=
+ 8 + 8 + 125 + 8 = "Sing = 5. 30

Supremum distance = Max (2 , 2 , S 2) = 5

,
data

Cluster
OutpuDie darnational
analysis
a dann cubes
Multidimensional data tables
Outlier
mining

Support (x = y) P(Xuy)
=

Confidence (x -y) P(Y(X)

= =

parallel and distributed data mining algorimms .

Loose
coupling
Semitight coupling
fight coupling
& IAP -- Online
analytical processing

Chapter & Data

processing
Data reduction

Data cleaning -
> Data Integration
Data transformation Q5-Q
Interquartile range (IQR)
- distance between ,

Distributive Measure
Algebraic Measure

weighted arimmatic
trimmed Mean
mean

I wi or weightedaverage

Holistic Measure

Median = L
+fat) wide

>
-
unimodal
& Sima
al
dispersion or variance

quartile I
as

like

first quartile Q
Boxplots for visualizing
Variance N X ,, Xz XN
of
. ...

= [Evi-Y(Exi)]
mean Value

Quantile plot

fi =

o 5
.

quantile-quantile plow or
giq put
Scatter
plot
Loes curve

data
smooring &
BinningMenodeion
Clustering

field overloading

uniqueeve re

entity identification problem

correlation analys is
Correlation
Chi
crefficient (Pearson's product moment) =

rAB = )

=
T
square
Contingency table

invert
Data traformation can

constrative
Win-max normalization transform
z-score normalization
/DAT) -
* worde
i
C
Decision
C1 5
tree
and Cort
distrate
wavelet transforms
ID3
:

Dimensionality Reduction - principal components analysis

Pyramid algorithm
Principal components analysis
 Info(D) = Entropy (D + Entropy (D

Entropy (D , ) =
= pilry(pi) .
3-1-S' rule

Large Software Can be

developed
by following

↑
Waterfall
Mush
a
Spin

~ 44)2
using 4432 + (23 - 46 .

+ 127 - 42 .

19511824
1476 1
78 +

44)" + (39-46 44)

-4 6 (41-16 443
. · .

121 54
.
-us . um)2+ 149-usun) + (so-usunsh

18 &

24
24706 gyz5 96 +1011
.

4477 22 .

-
min-Mad X = X-minx)
-
(manmin) + min

max(X) -
minD)

z-scoremeanStandard deviation
z-score normalization = Mean absolute deviation (MaD) =

↑ xi-y
mean
largest
number digic
Normalization decimal DX
by =
=/ 0
2x1000 = 3

equi-wide bin-min( new (min + u (eq)(

number
of bing
k-means- >
clustering
Given datax : S 10 11
, , ,
13 15, 35 , 50
, ,
55
; 72 , 92, 201, 215

Step 1 initial Gues

Take one number ative

beginning mid und
If 50 doo

Step
2 make a table and measure distance

X Distance to 10 Distance to 50 Distance to 200 Cluster Assignment

g S 45 195 Cluster I

10 8 48 198 Cluster 1

11 I 39 189 Cluster 1

13 3 37 187 Cluster I

Is S 35 189 Cluster 1
35 25 IS 16S Cluster z

So 40 J Iso cluster z

SS YS S 14S Cluster
72 62 22 128 Cluster z

92 82 42 10s Cluster 2

201 194 154 4 Cluster 3

215- 205 165 Is Cluster]

closter 1 :
,,
5 10 11 ,
13 , 15

Cluster 2 : 35, so 55 72
, , , 92

Cluster 3 : 201 His

,