# October 7, 2011 Data Nining: Concepts and Techniques 1

Data Nining:
Concepts and Techniques
- Chapter S -
]iawei Han
Department of Computer Science
University of !llinois at Urbana·Champaign
www.cs.uiuc.edu/~hanj
October 7, 2011 Data Nining: Concepts and Techniques 2
Chapter S: Nining Frequent Patterns,
Association and Correlations
Basic concepts and a road map
Efficient and scalable frequent itemset mining
methods
Nining various kinds of association rules
From association mining to correlation
analysis
Constraint·based association mining
Summary
October 7, 2011 Data Nining: Concepts and Techniques 3
What !s Frequent Pattern Analysis?
Frequent pattern: a pattern (a set of items, subsequences, substructures,
etc.) that occurs frequently in a data set
First proposed by Agrawal, !mielinski, and Swami |A!S33] in the context
of frequent itemsets and association rule mining
Notivation: Finding inherent regularities in data
What products were often purchased together?- Beer and diapers?!
What are the subsequent purchases after buying a PC?
What kinds of DNA are sensitive to this new drug?
Can we automatically classify web documents?
Applications
Basket data analysis, cross·marketing, catalog design, sale campaign
analysis, Web log (click stream) analysis, and DNA sequence analysis.
October 7, 2011 Data Nining: Concepts and Techniques 4
Why !s Freq. Pattern Nining !mportant?
Discloses an intrinsic and important property of data sets
Forms the foundation for many essential data mining tasks
Association, correlation, and causality analysis
Sequential, structural (e.g., sub·graph) patterns
Pattern analysis in spatiotemporal, multimedia, time·
series, and stream data
Classification: associative classification
Cluster analysis: frequent pattern·based clustering
Data warehousing: iceberg cube and cube·gradient
Semantic data compression: fascicles
October 7, 2011 Data Nining: Concepts and Techniques S
Basic Concepts: Frequent Patterns and
Association Rules
!temset X = {x
1
, ., x
k
}
Find all the rules X Y with minimum
support and confidence
support, s, probability that a
transaction contains X Y
confidence, c, conditional
probability that a transaction
having X also contains Y
Let sup
min
= S0%, conf
min
= S0%
Freq. Pat.: {A:3, B:3, D:4, E:3, AD:3}
Association rules:
A D (60%, 100%)
D A (60%, 7S%)
Customer
Customer
Customer
Transaction·id !tems bought
10 A, B, D
20 A, C, D
30 A, D, E
40 B, E, F
S0 B, C, D, E, F
October 7, 2011 Data Nining: Concepts and Techniques 6
Closed Patterns and Nax·Patterns
A long pattern contains a combinatorial number of sub·
patterns, e.g., {a
1
, ., a
100
} contains (
100
1
) + (
100
2
) + . +
(
1
1
0
0
0
0
) = 2
100
- 1 = 1.27*10
30
sub·patterns!
Solution: Nine closed patterns and max·patterns instead
An itemset X is closed if X is frequent and there exists no
super·pattern Y 5 X, with the same support as X
(proposed by Pasquier, et al. @ !CDT'33)
An itemset X is a max·pattern if X is frequent and there
exists no frequent super·pattern Y 5 X (proposed by
Bayardo @ S!CNOD'38)
Closed pattern is a lossless compression of freq. patterns
Reducing the # of patterns and rules
October 7, 2011 Data Nining: Concepts and Techniques 7
Closed Patterns and Nax·Patterns
Exercise. DB = {<a
1
, ., a
100
>, < a
1
, ., a
S0
>}
Nin_sup = 1.
What is the set of closed itemset?
<a
1
, ., a
100
>: 1
< a
1
, ., a
S0
>: 2
What is the set of max·pattern?
<a
1
, ., a
100
>: 1
What is the set of all patterns?
!!
October 7, 2011 Data Nining: Concepts and Techniques 8
Chapter S: Nining Frequent Patterns,
Association and Correlations
Basic concepts and a road map
Efficient and scalable frequent itemset mining
methods
Nining various kinds of association rules
From association mining to correlation
analysis
Constraint·based association mining
Summary
October 7, 2011 Data Nining: Concepts and Techniques 3
Scalable Nethods for Nining Frequent Patterns
The downward closure property of frequent patterns
Any subset of a frequent itemset must be frequent
!f <beer, diaper, nuts} is frequent, so is <beer,
diaper}
i.e., every transaction having {beer, diaper, nuts} also
contains {beer, diaper}
Scalable mining methods: Three major approaches
Apriori (Agrawal S Srikant@vLDB'34)
Freq. pattern growth (FPgrowth-Han, Pei S Yin
@S!CNOD'00)
vertical data format approach (Charm-Zaki S Hsiao
@SDN'02)
October 7, 2011 Data Nining: Concepts and Techniques 10
Apriori: A Candidate Ceneration·and·Test Approach
Apriori pruning principle: !f there is any itemset which is
infrequent, its superset should not be generated/tested!
(Agrawal S Srikant @vLDB'34, Nannila, et al. @ KDD' 34)
Nethod:
!nitially, scan DB once to get frequent 1·itemset
Cenerate length (k+1) candidate itemsets from length k
frequent itemsets
Test the candidates against DB
Terminate when no frequent or candidate set can be
generated
October 7, 2011 Data Nining: Concepts and Techniques 11
The Apriori Algorithm-An Example
Database TDB
1
st
scan
C
1
L
1
L
2
C
2
C
2

nd
scan
C
3
L
3
3
rd
scan
Tid !tems
10 A, C, D
20 B, C, E
30 A, B, C, E
40 B, E
!temset sup
{A} 2
{B} 3
{C} 3
{D} 1
{E} 3
!temset sup
{A} 2
{B} 3
{C} 3
{E} 3
!temset
{A, B}
{A, C}
{A, E}
{B, C}
{B, E}
{C, E}
!temset sup
{A, B} 1
{A, C} 2
{A, E} 1
{B, C} 2
{B, E} 3
{C, E} 2
!temset sup
{A, C} 2
{B, C} 2
{B, E} 3
{C, E} 2
!temset
{B, C, E}
!temset sup
{B, C, E} 2
Sup
min
= 2
October 7, 2011 Data Nining: Concepts and Techniques 12
The Apriori Algorithm
Pseudo·code:
C
k
: Candidate itemset of size k
L
k
: frequent itemset of size k
L
1
= {frequent items},
or (k = 1, L
k
C
k+1
= candidates generated from L
k
,
or each transaction t in database do
increment the count of all candidates in C
k+1
that are contained in t
L
k+1
= candidates in C
k+1
with min_support
end
return
k
L
k
,
October 7, 2011 Data Nining: Concepts and Techniques 13
!mportant Details of Apriori
How to generate candidates?
Step 1: self·joining L
k
Step 2: pruning
How to count supports of candidates?
Example of Candidate·generation
L
3
={abc, abd, acd, ace, bcd}
Self·joining: L
3
*L
3
abcd from abc and abd
acde from acd and ace
Pruning:
acde is removed because ade is not in L
3
C
4
={abcd}
October 7, 2011 Data Nining: Concepts and Techniques 14
How to Cenerate Candidates?
Suppose the items in L
k·1
are listed in an order
Step 1: self·joining L
k·1
insert into C
k
select p.item
1
, p.item
2
, ., p.item
k·1
, q.item
k·1
from L
k·1
p, L
k·1
q
where p.item
1
=q.item
1
, ., p.item
k·2
=q.item
k·2
, p.item
k·1
<
q.item
k·1
Step 2: pruning
forall itemsets c in C
k
do
forall (k·1)·subsets s o c do
i (s is not in L
k·1
) then delete c rom C
k
October 7, 2011 Data Nining: Concepts and Techniques 1S
How to Count Supports of Candidates?
Why counting supports of candidates a problem?
The total number of candidates can be very huge
One transaction may contain many candidates
Nethod:
Candidate itemsets are stored in a hash·tree
Leaf node of hash·tree contains a list of itemsets and
counts
!nterior node contains a hash table
Subset function: finds all the candidates contained in
a transaction
October 7, 2011 Data Nining: Concepts and Techniques 16
Example: Counting Supports of Candidates
1,4,7
,5,8
3,6,9
Subset Iunction
3 4
5 6 7
1 4 5
1 3 6
1 4
4 5 7
1 5
4 5 8
1 5 9
3 4 5 3 5 6
3 5 7
6 8 9
3 6 7
3 6 8
Transaction: 1 3 5 6
1 ¹ 3 5 6
1 ¹ 3 5 6
1 3 ¹ 5 6
October 7, 2011 Data Nining: Concepts and Techniques 17
Efficient !mplementation of Apriori in SOL
Hard to get good performance out of pure SOL (SOL·
32) based approaches alone
Nake use of object·relational extensions like UDFs,
BLOBs, Table functions etc.
Cet orders of magnitude improvement
S. Sarawagi, S. Thomas, and R. Agrawal. !ntegrating
association rule mining with relational database
systems: Alternatives and implications. !n S!CNOD'38
October 7, 2011 Data Nining: Concepts and Techniques 18
Challenges of Frequent Pattern Nining
Challenges
Nultiple scans of transaction database
Huge number of candidates
Tedious workload of support counting for candidates
!mproving Apriori: general ideas
Reduce passes of transaction database scans
Shrink number of candidates
Facilitate support counting of candidates
October 7, 2011 Data Nining: Concepts and Techniques 13
Partition: Scan Database Only Twice
Any itemset that is potentially frequent in DB must be
frequent in at least one of the partitions of DB
Scan 1: partition database and find local frequent
patterns
Scan 2: consolidate global frequent patterns
A. Savasere, E. Omiecinski, and S. Navathe. An efficient
algorithm for mining association in large databases. !n
vLDB'3S
October 7, 2011 Data Nining: Concepts and Techniques 20
DHP: Reduce the Number of Candidates
A k·itemset whose corresponding hashing bucket count is
below the threshold cannot be frequent
Candidates: a, b, c, d, e
Hash entries: {ab, ad, ae} {bd, be, de} .
Frequent 1·itemset: a, b, d, e
ab is not a candidate 2·itemset if the sum of count of
{ab, ad, ae} is below support threshold
]. Park, N. Chen, and P. Yu. An effective hash·based
algorithm for mining association rules. !n S!CNOD'3S
October 7, 2011 Data Nining: Concepts and Techniques 21
Sampling for Frequent Patterns
Select a sample of original database, mine frequent
patterns within sample using Apriori
Scan database once to verify frequent itemsets found in
sample, only borders of closure of frequent patterns are
checked
Example: check abcd instead of ab, ac, ., etc.
Scan database again to find missed frequent patterns
H. Toivonen. Sampling large databases for association
rules. !n vLDB'36
October 7, 2011 Data Nining: Concepts and Techniques 22
D!C: Reduce Number of Scans
BCD
BC BD CD BCD
B C BC D BD CD
B C D
¦}
Itemset lattice
Once both A and D are determined
frequent, the counting of AD begins
Once all length·2 subsets of BCD are
determined frequent, the counting of BCD
begins
Transactions
1-itemsets
-itemsets
.
5riori
1-itemsets
-items
3-items DIC
S. Brin R. Notwani, ]. Ullman,
and S. Tsur. Dynamic itemset
counting and implication rules for
S!CNOD'37
October 7, 2011 Data Nining: Concepts and Techniques 23
Bottleneck of Frequent·pattern Nining
Nultiple database scans are costly
Nining long patterns needs many passes of
scanning and generates lots of candidates
To find frequent itemset i
1
i
2
.i
100
# of scans: 100
# of Candidates: (
100
1
) + (
100
2
) + . + (
1
1
0
0
0
0
) = 2
100
·
1 = 1.27*10
30
!
Bottleneck: candidate·generation·and·test
Can we avoid candidate generation?
October 7, 2011 Data Nining: Concepts and Techniques 24
Nining Frequent Patterns Without
Candidate Ceneration
Crow long patterns from short ones using local
frequent items
¨abc" is a frequent pattern
Cet all transactions having ¨abc": DB]abc
¨d" is a local frequent item in DB]abc abcd is a
frequent pattern
October 7, 2011 Data Nining: Concepts and Techniques 2S
Construct FP·tree from a Transaction Database
¦}
f.4 c.1
b.1
p.1
b.1 c.3
a.3
b.1 m.2
p.2 m.1
f 4
c 4
a 3
b 3
m 3
p 3
min¸support ÷ 3
1ID Items bought (ordered) frequent items
100 ]f, a, c, d, g, i, m, p] ]f, c, a, m, p]
200 ]a, b, c, f, l, m, o] ]f, c, a, b, m]
300 ]b, f, h, j, o, w] ]f, b]
400 ]b, c, k, s, p] ]c, b, p]
500 ]a, f, c, e, l, p, m, n] ]f, c, a, m, p]
1. Scan DB once, find
frequent 1·itemset
(single item pattern)
2. Sort frequent items in
frequency descending
order, f·list
3. Scan DB again,
construct FP·tree
F·list=f·c·a·b·m·p
October 7, 2011 Data Nining: Concepts and Techniques 26
Benefits of the FP·tree Structure
Completeness
Preserve complete information for frequent pattern
mining
Never break a long pattern of any transaction
Compactness
Reduce irrelevant info-infrequent items are gone
!tems in frequency descending order: the more
frequently occurring, the more likely to be shared
Never be larger than the original database (not count
For Connect·4 DB, compression ratio could be over 100
October 7, 2011 Data Nining: Concepts and Techniques 27
Partition Patterns and Databases
Frequent patterns can be partitioned into subsets
according to f·list
F·list=f·c·a·b·m·p
Patterns containing p
Patterns having m but no p
.
Patterns having c but no a nor b, m, p
Pattern f
Completeness and non·redundency
October 7, 2011 Data Nining: Concepts and Techniques 28
Find Patterns Having P From P·conditional Database
Starting at the frequent item header table in the FP·tree
Traverse the FP·tree by following the link of each frequent item p
Accumulate all of transformed prefix paths of item p to form p's
conditional pattern base
onditional pattern bases
item cond. pattern base
c f:3
a fc:3
b fca:1, f:1, c:1
m fca:2, fcab:1
p fcam:2, cb:1
¦}
f.4 c.1
b.1
p.1
b.1 c.3
a.3
b.1 m.2
p.2 m.1
f 4
c 4
a 3
b 3
m 3
p 3
October 7, 2011 Data Nining: Concepts and Techniques 23
From Conditional Pattern·bases to Conditional FP·trees
For each pattern·base
Accumulate the count for each item in the base
Construct the FP·tree for the frequent items of the
pattern base
m-conditional pattern base:
fca:2, fcab:1
¦}
f.3
c.3
a.3
m-conditional !tree
All frequent
patterns relate to m
m,
fm, cm, am,
fcm, fam, cam,
fcam

¦}
f.4 c.1
b.1
p.1
b.1 c.3
a.3
b.1 m.2
p.2 m.1
f 4
c 4
a 3
b 3
m 3
p 3
October 7, 2011 Data Nining: Concepts and Techniques 30
Recursion: Nining Each Conditional FP·tree
¦}
f.3
c.3
a.3
m-conditional !tree
Cond. pattern base of ¨am": (fc:3)
¦}
f.3
c.3
am-conditional !tree
Cond. pattern base of ¨cm": (f:3)
¦}
f.3
cm-conditional !tree
Cond. pattern base of ¨cam": (f:3)
¦}
f.3
cam-conditional !tree
October 7, 2011 Data Nining: Concepts and Techniques 31
A Special Case: Single Prefix Path in FP·tree
Suppose a (conditional) FP·tree T has a shared
single prefix·path P
Nining can be decomposed into two parts
Reduction of the single prefix path into one node
Concatenation of the mining results of the two
parts

a
2
.n
2
a
3
.n
3
a
1
.n
1
¦}
b
1
:m
1
C
1
:k
1
C
2
:k
2
C
3
:k
3
b
1
:m
1
C
1
:k
1
C
2
:k
2
C
3
:k
3
r
1

a
2
.n
2
a
3
.n
3
a
1
.n
1
¦}
r
1

October 7, 2011 Data Nining: Concepts and Techniques 32
Nining Frequent Patterns With FP·trees
!dea: Frequent pattern growth
Recursively grow frequent patterns by pattern and
database partition
Nethod
For each frequent item, construct its conditional
pattern·base, and then its conditional FP·tree
Repeat the process on each newly created conditional
FP·tree
Until the resulting FP·tree is empty, or it contains only
one path-single path will generate all the
combinations of its sub·paths, each of which is a
frequent pattern
October 7, 2011 Data Nining: Concepts and Techniques 33
Scaling FP·growth by DB Projection
FP·tree cannot fit in memory?-DB projection
First partition a database into a set of projected DBs
Then construct and mine FP·tree for each projected DB
Parallel projection vs. Partition projection techniques
Parallel projection is space costly
October 7, 2011 Data Nining: Concepts and Techniques 34
Partition·based Projection
Parallel projection needs a lot
of disk space
Partition projection saves it
Tran. DB
fcamp
fcabm
fb
cbp
fcamp
p·proj DB
fcam
cb
fcam
m·proj DB
fcab
fca
fca
b·proj DB
f
cb
.
a·proj DB
fc
.
c·proj DB
f
.
·proj DB
.
am·proj DB
fc
fc
fc
cm·proj DB
f
f
f
.
October 7, 2011 Data Nining: Concepts and Techniques 3S
FP·Crowth vs. Apriori: Scalability With the Support
Threshold
0
10
20
30
40
50
60
70
80
90
100
0 0.5 1 1.5 2 2.5 3
Support threshoId(%)
R
u
n

t
i
m
e
(
s
e
c
.
)
1 FP-growth runtime
1 Apriori runtime
Data set T5ID1
October 7, 2011 Data Nining: Concepts and Techniques 36
FP·Crowth vs. Tree·Projection: Scalability with
the Support Threshold
0
20
40
60
80
100
120
140
0 0.5 1 1.5 2
Support threshoId (%)
R
u
n
t
i
m
e

(
s
e
c
.
)
2 FP-growth
2 TreeProjection
Data set T5ID1
October 7, 2011 Data Nining: Concepts and Techniques 37
Why !s FP·Crowth the Winner?
Divide·and·conquer:
decompose both the mining task and DB according to
the frequent patterns obtained so far
leads to focused search of smaller databases
Other factors
no candidate generation, no candidate test
compressed database: FP·tree structure
no repeated scan of entire database
basic ops-counting local freq items and building sub
FP·tree, no pattern search and matching
October 7, 2011 Data Nining: Concepts and Techniques 38
!mplications of the Nethodology
Nining closed frequent itemsets and max·patterns
CLOSET (DNKD'00)
Nining sequential patterns
FreeSpan (KDD'00), PrefixSpan (!CDE'01)
Constraint·based mining of frequent patterns
Convertible constraints (KDD'00, !CDE'01)
Computing iceberg data cubes with complex measures
H·tree and H·cubing algorithm (S!CNOD'01)
October 7, 2011 Data Nining: Concepts and Techniques 33
NaxNiner: Nining Nax·patterns
1
st
scan: find frequent items
A, B, C, D, E
2
nd
scan: find support for
BC, BD, BE, BCDE
CD, CE, CDE, DE,
Since BCDE is a max·pattern, no need to check BCD, BDE,
CDE in later scan
R. Bayardo. Efficiently mining long patterns from
databases. !n S!CNOD'38
Tid !tems
10 A,B,C,D,E
20 B,C,D,E,
30 A,C,D,F
Potential
max·patterns
October 7, 2011 Data Nining: Concepts and Techniques 40
Nining Frequent Closed Patterns: CLOSET
Flist: list of all frequent items in support ascending order
Flist: d·a·f·e·c
Divide search space
Patterns having d
Patterns having d but no a, etc.
Find frequent closed pattern recursively
Every transaction having d also has cfa cfad is a
frequent closed pattern
]. Pei, ]. Han S R. Nao. CLOSET: An Efficient Algorithm for
Nining Frequent Closed !temsets", DNKD'00.
T!D !tems
10 a, c, d, e, f
20 a, b, e
30 c, e, f
40 a, c, d, f
S0 c, e, f
Nin_sup=2
October 7, 2011 Data Nining: Concepts and Techniques 41
CLOSET+: Nining Closed !temsets by
Pattern·Crowth
!temset merging: if Y appears in every occurrence of X, then Y
is merged with X
Sub·itemset pruning: if Y 5 X, and sup(X) = sup(Y), X and all of
X's descendants in the set enumeration tree can be pruned
Hybrid tree projection
Bottom·up physical tree·projection
Top·down pseudo tree·projection
!tem skipping: if a local frequent item has the same support in
several header tables at different levels, one can prune it from
the header table at higher levels
Efficient subset checking
October 7, 2011 Data Nining: Concepts and Techniques 42
CHARN: Nining by Exploring vertical Data Format
vertical format: t(AB) = {T
11
, T
2S
, .}
tid·list: list of trans.·ids containing an itemset
Deriving closed patterns based on vertical intersections
t(X) = t(Y): X and Y always happen together
t(X) · t(Y): transaction having X always has Y
Using diffset to accelerate mining
Only keep track of differences of tids
t(X) = {T
1
, T
2
, T
3
}, t(XY) = {T
1
, T
3
}
Diffset (XY, X) = {T
2
}
Eclat/NaxEclat (Zaki et al. @KDD'37), v!PER(P. Shenoy et
al.@S!CNOD'00), CHARN (Zaki S Hsiao@SDN'02)
October 7, 2011 Data Nining: Concepts and Techniques 43
Further !mprovements of Nining Nethods
AFOPT (Liu, et al. @ KDD'03)
A ¨push·right" method for mining condensed frequent
pattern (CFP) tree
Carpenter (Pan, et al. @ KDD'03)
Nine data sets with small rows but numerous columns
Construct a row·enumeration tree for efficient mining
October 7, 2011 Data Nining: Concepts and Techniques 44
visualization of Association Rules: Plane Craph
October 7, 2011 Data Nining: Concepts and Techniques 4S
visualization of Association Rules: Rule Craph
October 7, 2011 Data Nining: Concepts and Techniques 46
visualization of Association Rules
(SC!/NineSet 3.0)
October 7, 2011 Data Nining: Concepts and Techniques 47
Chapter S: Nining Frequent Patterns,
Association and Correlations
Basic concepts and a road map
Efficient and scalable frequent itemset mining
methods
Nining various kinds of association rules
From association mining to correlation
analysis
Constraint·based association mining
Summary
October 7, 2011 Data Nining: Concepts and Techniques 48
Nining various Kinds of Association Rules
Nining multilevel association
Niming multidimensional association
Nining quantitative association
Nining interesting correlation patterns
October 7, 2011 Data Nining: Concepts and Techniques 43
Nining Nultiple·Level Association Rules
!tems often form hierarchies
Flexible support settings
!tems at the lower level are expected to have lower
support
Exploration of shared multi·level mining (Agrawal S
Srikant@vLB'3S, Han S Fu@vLDB'3S)
uniform support
ilk
[support 10º]
2º ilk
[support 6º]
Skim ilk
[support 4º]
Level 1
min¸sup 5º
Level 2
min¸sup 5º
Level 1
min¸sup 5º
Level 2
min¸sup 3º
reduced support
October 7, 2011 Data Nining: Concepts and Techniques S0
Nulti·level Association: Redundancy Filtering
Some rules may be redundant due to ¨ancestor"
relationships between items.
Example
milk ÷ wheat bread |support = 8%, confidence = 70%]
2% milk ÷ wheat bread |support = 2%, confidence = 72%]
We say the first rule is an ancestor of the second rule.
A rule is redundant if its support is close to the ¨expected"
value, based on the rule's ancestor.
October 7, 2011 Data Nining: Concepts and Techniques S1
Nining Nulti·Dimensional Association
Single·dimensional rules:
Nulti·dimensional rules: > 2 dimensions or predicates
!nter·dimension assoc. rules (no repeated predicates)
age(X,"13·2S") · occupation(X,¨student") ÷ buys(X, ¨coke")
hybrid·dimension assoc. rules (repeated predicates)
Categorical Attributes: finite number of possible values, no
ordering among values-data cube approach
Ouantitative Attributes: numeric, implicit ordering among
October 7, 2011 Data Nining: Concepts and Techniques S2
Nining Ouantitative Associations
Techniques can be categorized by how numerical
attributes, such as age or salary are treated
1. Static discretization based on predefined concept
hierarchies (data cube methods)
2. Dynamic discretization based on data distribution
(quantitative rules, e.g., Agrawal S Srikant@S!CNOD36)
3. Clustering: Distance·based association (e.g., Yang S
Niller@S!CNOD37)
one dimensional clustering then association
4. Deviation: (such as Aumann and Lindell@KDD33)
Sex = female => Wage: mean=\$7/hr (overall mean = \$3)
October 7, 2011 Data Nining: Concepts and Techniques S3
Static Discretization of Ouantitative Attributes
Discretized prior to mining using concept hierarchy.
Numeric values are replaced by ranges.
!n relational database, finding all frequent k·predicate sets
will require k or k+1 table scans.
Data cube is well suited for mining.
The cells of an n·dimensional
cuboid correspond to the
predicate sets.
Nining from data cubes
can be much faster.
income) age)
)
October 7, 2011 Data Nining: Concepts and Techniques S4
Ouantitative Association Rules
age(X.¨3435¨) · income(X.¨3050K¨)
Proposed by Lent, Swami and Widom !CDE'37
Numeric attributes are dynamically discretized
Such that the confidence or compactness of the rules
mined is maximized
2·D quantitative association rules: A
quan1
· A
quan2
÷ A
cat
association rules
to form general
rules using a 2·D grid
Example
October 7, 2011 Data Nining: Concepts and Techniques SS
Nining Other !nteresting Patterns
Flexible support constraints (Wang et al. @ vLDB'02)
Some items (e.g., diamond) may occur rarely but are
valuable
Customized sup
min
specification and application
Top·K closed frequent patterns (Han, et al. @ !CDN'02)
Hard to specify sup
min
, but top·k with length
min
is more
desirable
Dynamically raise sup
min
in FP·tree construction and
mining, and select most promising path to mine
October 7, 2011 Data Nining: Concepts and Techniques S6
Chapter S: Nining Frequent Patterns,
Association and Correlations
Basic concepts and a road map
Efficient and scalable frequent itemset mining
methods
Nining various kinds of association rules
From association mining to correlation analysis
Constraint·based association mining
Summary
October 7, 2011 Data Nining: Concepts and Techniques S7
!nterestingness Neasure: Correlations (Lift)
The overall % of students eating cereal is 7S% > 66.7%.
play basketball ÷ not eat cereal |20%, 33.3%] is more accurate,
although with lower support and confidence
Neasure of dependent/correlated events: lift
89 .
5 375 5 3
5
) , = = C B lift
Cereal 2000 17S0 37S0
Not cereal 1000 2S0 12S0
Sum(col.) 3000 2000 S000
) )
)
B P A P
B A P
lift

=
33 . 1
5 15 5 3
5 1
) , = = C B lift
October 7, 2011 Data Nining: Concepts and Techniques S8
Are lift and /
2
Cood Neasures of Correlation?
if 8S% of customers buy milk
Support and confidence are not good to represent correlations
So many interestingness measures? (Tan, Kumar, Sritastava @KDD'02)
Nilk No Nilk Sum (row)
Coffee m, c ~m, c c
No Coffee m, ~c ~m, ~c ~c
Sum(col.) m ~m %
) )
)
B P A P
B A P
lift

=
DB m, c ~m, c m~c ~m~c lift all·conf coh /2
A1 1000 100 100 10,000 3.26 0.31 0.83 30SS
A2 100 1000 1000 100,000 8.44 0.03 0.0S 670
A3 1000 100 10000 100,000 3.18 0.03 0.03 8172
A4 1000 1000 1000 1000 1 0.S 0.33 0
) su5 * max*
) su5
*
item

conf all =
)
) su5
:niverse

coh =
October 7, 2011 Data Nining: Concepts and Techniques S3
Which Neasures Should Be Used?
lit and /
2
are not
good measures for
correlations in large
transactional DBs
all·con or
coherence could be
good measures
(Omiecinski@TKDE'03)
Both all·con and
coherence have the
downward closure
property
Efficient algorithms
can be derived for
mining (Lee et al.
@!CDN'03sub)
October 7, 2011 Data Nining: Concepts and Techniques 60
Chapter S: Nining Frequent Patterns,
Association and Correlations
Basic concepts and a road map
Efficient and scalable frequent itemset mining
methods
Nining various kinds of association rules
From association mining to correlation analysis
Constraint·based association mining
Summary
October 7, 2011 Data Nining: Concepts and Techniques 61
Constraint·based (Ouery·Directed) Nining
Finding all the patterns in a database autonomously? -
unrealistic!
The patterns could be too many but not focused!
Data mining should be an interactive process
User directs what to be mined using a data mining
query language (or a graphical user interface)
Constraint·based mining
User flexibility: provides constraints on what to be
mined
System optimization: explores such constraints for
efficient mining-constraint·based mining
October 7, 2011 Data Nining: Concepts and Techniques 62
Constraints in Data Nining
Knowledge type constraint:
classification, association, etc.
Data constraint - using SOL·like queries
find product pairs sold together in stores in Chicago in
Dec.'02
Dimension/level constraint
in relevance to region, price, brand, customer category
Rule (or pattern) constraint
small sales (price < \$10) triggers big sales (sum >
\$200)
!nterestingness constraint
strong rules: min_support > 3%, min_confidence >
60%
October 7, 2011 Data Nining: Concepts and Techniques 63
Constrained Nining vs. Constraint·Based Search
Constrained mining vs. constraint·based search/reasoning
Both are aimed at reducing search space
Finding all patterns satisfying constraints vs. finding
some (or one) answer in constraint·based search in A!
Constraint·pushing vs. heuristic search
!t is an interesting research problem on how to integrate
them
Constrained mining vs. query processing in DBNS
Database query processing requires to find all
Constrained pattern mining shares a similar philosophy
as pushing selections deeply in query processing
October 7, 2011 Data Nining: Concepts and Techniques 64
Anti·Nonotonicity in Constraint Pushing
Anti·monotonicity
When an intemset S iolates the
constraint, so does any of its superset
sum(S.Price) A v is anti·monotone
sum(S.Price) > v is not anti·monotone
Example. C: range(S.profit) A 1S is anti·
monotone
!temset ab violates C
So does every superset of ab
T!D Transaction
10 a, b, c, d, f
20 b, c, d, f, g, h
30 a, c, d, e, f
40 c, e, f, g
TDB min*su5)
!tem Proit
a 40
b 0
c ·20
d 10
e ·30
f 30
g 20
h ·10
October 7, 2011 Data Nining: Concepts and Techniques 6S
Nonotonicity for Constraint Pushing
Nonotonicity
When an intemset S satisies the
constraint, so does any of its
superset
sum(S.Price) > v is monotone
min(S.Price) A v is monotone
Example. C: range(S.profit) > 1S
!temset ab satisfies C
So does every superset of ab
T!D Transaction
10 a, b, c, d, f
20 b, c, d, f, g, h
30 a, c, d, e, f
40 c, e, f, g
TDB min*su5)
!tem Proit
a 40
b 0
c ·20
d 10
e ·30
f 30
g 20
h ·10
October 7, 2011 Data Nining: Concepts and Techniques 66
Succinctness
Succinctness:
Civen A
1,
the set of items satisfying a succinctness
constraint C, then any set S satisfying C is based on
A
1
, i.e., S contains a subset belonging to A
1
!dea: Without looking at the transaction database,
whether an itemset S satisfies constraint C can be
determined based on the selection of items
min(S.Price) A v is succinct
sum(S.Price) > v is not succinct
Optimization: !f C is succinct, C is pre·counting pushable
October 7, 2011 Data Nining: Concepts and Techniques 67
The Apriori Algorithm - Example
TÌ Ìtems
100 1 3 4
200 2 3 5
300 1 2 3 5
400 2 5
Database D itemset sup.
{1} 2
{2} 3
{3} 3
{4} 1
{5} 3
itemset sup.
{1} 2
{2} 3
{3} 3
{5} 3
Scan D
C
1
L
1
itemset
{1 2}
{1 3}
{1 5}
{2 3}
{2 5}
{3 5}
itemset sup
{1 2} 1
{1 3} 2
{1 5} 1
{2 3} 2
{2 5} 3
{3 5} 2
itemset sup
{1 3} 2
{2 3} 2
{2 5} 3
{3 5} 2
L
2
C
2
C
2
Scan D
C
3
L
3
itemset
{2 3 5}
Scan D
itemset sup
{2 3 5} 2
October 7, 2011 Data Nining: Concepts and Techniques 68
Naive Algorithm: Apriori + Constraint
TÌ Ìtems
100 1 3 4
200 2 3 5
300 1 2 3 5
400 2 5
Database D itemset sup.
{1} 2
{2} 3
{3} 3
{4} 1
{5} 3
itemset sup.
{1} 2
{2} 3
{3} 3
{5} 3
Scan D
C
1
L
1
itemset
{1 2}
{1 3}
{1 5}
{2 3}
{2 5}
{3 5}
itemset sup
{1 2} 1
{1 3} 2
{1 5} 1
{2 3} 2
{2 5} 3
{3 5} 2
itemset sup
{1 3} 2
{2 3} 2
{2 5} 3
{3 5} 2
L
2
C
2
C
2
Scan D
C
3
L
3
itemset
{2 3 5}
Scan D
itemset sup
{2 3 5} 2
Constraint:
Sum<S.price} < S
October 7, 2011 Data Nining: Concepts and Techniques 63
The Constrained Apriori Algorithm: Push
an Anti·monotone Constraint Deep
TÌ Ìtems
100 1 3 4
200 2 3 5
300 1 2 3 5
400 2 5
Database D itemset sup.
{1} 2
{2} 3
{3} 3
{4} 1
{5} 3
itemset sup.
{1} 2
{2} 3
{3} 3
{5} 3
Scan D
C
1
L
1
itemset
{1 2}
{1 3}
{1 5}
{2 3}
{2 5}
{3 5}
itemset sup
{1 2} 1
{1 3} 2
{1 5} 1
{2 3} 2
{2 5} 3
{3 5} 2
itemset sup
{1 3} 2
{2 3} 2
{2 5} 3
{3 5} 2
L
2
C
2
C
2
Scan D
C
3
L
3
itemset
{2 3 5}
Scan D
itemset sup
{2 3 5} 2
Constraint:
Sum<S.price} < S
October 7, 2011 Data Nining: Concepts and Techniques 70
The Constrained Apriori Algorithm: Push a
Succinct Constraint Deep
TÌ Ìtems
100 1 3 4
200 2 3 5
300 1 2 3 5
400 2 5
Database D itemset sup.
{1} 2
{2} 3
{3} 3
{4} 1
{5} 3
itemset sup.
{1} 2
{2} 3
{3} 3
{5} 3
Scan D
C
1
L
1
itemset
{1 2}
{1 3}
{1 5}
{2 3}
{2 5}
{3 5}
itemset sup
{1 2} 1
{1 3} 2
{1 5} 1
{2 3} 2
{2 5} 3
{3 5} 2
itemset sup
{1 3} 2
{2 3} 2
{2 5} 3
{3 5} 2
L
2
C
2
C
2
Scan D
C
3
L
3
itemset
{2 3 5}
Scan D
itemset sup
{2 3 5} 2
Constraint:
min<S.price } <= 1
not immediately
to be used
October 7, 2011 Data Nining: Concepts and Techniques 71
Converting ¨Tough" Constraints
Convert tough constraints into anti·
monotone or monotone by properly
ordering items
Examine C: avg(S.profit) > 2S
Order items in value·descending
order
<a, f, g, d, b, h, c, e>
!f an itemset afb violates C
So does afbh, afb*
!t becomes anti·monotone!
T!D Transaction
10 a, b, c, d, f
20 b, c, d, f, g, h
30 a, c, d, e, f
40 c, e, f, g
TDB min*su5)
!tem Proit
a 40
b 0
c ·20
d 10
e ·30
f 30
g 20
h ·10
October 7, 2011 Data Nining: Concepts and Techniques 72
Strongly Convertible Constraints
avg(X) > 2S is convertible anti·monotone w.r.t.
item value descending order R: <a, f, g, d, b,
h, c, e>
!f an itemset af violates a constraint C, so
does every itemset with af as prefix, such as
afd
avg(X) > 2S is convertible monotone w.r.t. item
value ascending order R
·1
: <e, c, h, b, d, g, f,
a>
!f an itemset d satisfies a constraint C, so
does itemsets df and dfa, which having d as
a prefix
Thus, avg(X) > 2S is strongly convertible
!tem Proit
a 40
b 0
c ·20
d 10
e ·30
f 30
g 20
h ·10
October 7, 2011 Data Nining: Concepts and Techniques 73
Can Apriori Handle Convertible Constraint?
A convertible, not monotone nor anti·monotone
nor succinct constraint cannot be pushed deep
into the an Apriori mining algorithm
Within the level wise framework, no direct
pruning based on the constraint can be made
!temset df violates constraint C: avg(X)>=2S
Since adf satisfies C, Apriori needs df to
assemble adf, df cannot be pruned
But it can be pushed into frequent·pattern
growth framework!
!tem value
a 40
b 0
c ·20
d 10
e ·30
f 30
g 20
h ·10
October 7, 2011 Data Nining: Concepts and Techniques 74
Nining With Convertible Constraints
C: avg(X) >= 2S, min_sup=2
List items in every transaction in value descending
order R: <a, f, g, d, b, h, c, e>
C is convertible anti·monotone w.r.t. R
Scan TDB once
remove infrequent items
!tem h is dropped
!temsets a and f are good, .
Projection·based mining
!mposing an appropriate order on item projection
Nany tough constraints can be converted into
(anti)·monotone
T!D Transaction
10 a, f, d, b, c
20 f, g, d, b, c
30 a, f, d, c, e
40 f, g, h, c, e
TDB min*su5)
!tem value
a 40
f 30
g 20
d 10
b 0
h ·10
c ·20
e ·30
October 7, 2011 Data Nining: Concepts and Techniques 7S
Handling Nultiple Constraints
Different constraints may require different or even
conflicting item·ordering
!f there exists an order R s.t. both C
1
and C
2
are
convertible w.r.t. R, then there is no conflict between
the two convertible constraints
!f there exists conflict on order of items
Try to satisfy one constraint first
Then using the order for the other constraint to
mine frequent itemsets in the corresponding
projected database
October 7, 2011 Data Nining: Concepts and Techniques 76
What Constraints Are Convertible?
Constraint
Convertible anti·
monotone
Convertible
monotone
Strongly
convertible
avg(S) A , > v Yes Yes Yes
median(S) A , > v Yes Yes Yes
sum(S) A v (items could be of any value,
v > 0)
Yes No No
sum(S) A v (items could be of any value,
v A 0)
No Yes No
sum(S) > v (items could be of any value,
v > 0)
No Yes No
sum(S) > v (items could be of any value,
v A 0)
Yes No No
..
October 7, 2011 Data Nining: Concepts and Techniques 77
Constraint·Based Nining-A Ceneral Picture
Constraint Antimonotone Nonotone Succinct
Z S no yes yes
S v no yes yes
S v yes no yes
min(S) A no yes yes
min(S) > yes no yes
max(S) A yes no yes
max(S) > no yes yes
count(S) A yes no weakly
count(S) > no yes weakly
sum(S) A ( a Z S, a > 0 ) yes no no
sum(S) > ( a Z S, a > 0 ) no yes no
range(S) A yes no no
range(S) > no yes no
ag(S) o , o Z < =, A, > } conertible conertible no
support(S) > : yes no no
support(S) A : no yes no
October 7, 2011 Data Nining: Concepts and Techniques 78
A Classification of Constraints
Conertible
anti·monotone
Conertible
monotone
Strongly
conertible
!nconertible
Succinct
Antimonotone
Nonotone
October 7, 2011 Data Nining: Concepts and Techniques 73
Chapter S: Nining Frequent Patterns,
Association and Correlations
Basic concepts and a road map
Efficient and scalable frequent itemset mining
methods
Nining various kinds of association rules
From association mining to correlation analysis
Constraint·based association mining
Summary
October 7, 2011 Data Nining: Concepts and Techniques 80
Frequent·Pattern Nining: Summary
Frequent pattern mining-an important task in data mining
Scalable frequent pattern mining methods
Apriori (Candidate generation S test)
Projection·based (FPgrowth, CLOSET+, ...)
vertical format approach (CHARN, ...)
Nining a variety of rules and interesting patterns
Constraint·based mining
Nining sequential and structured patterns
Extensions and applications
October 7, 2011 Data Nining: Concepts and Techniques 81
Frequent·Pattern Nining: Research Problems
Nining fault·tolerant frequent, sequential and structured
patterns
Patterns allows limited faults (insertion, deletion,
mutation)
Nining truly interesting patterns
Surprising, novel, concise, .
Application exploration
E.g., DNA sequence analysis and bio·pattern
classification
¨!nvisible" data mining
October 7, 2011 Data Nining: Concepts and Techniques 82
Ref: Basic Concepts of Frequent Pattern Nining
(Association Rules) R. Agrawal, T. !mielinski, and A. Swami. Nining
association rules between sets of items in large databases.
S!CNOD'33.
(Nax·pattern) R. ]. Bayardo. Efficiently mining long patterns from
databases. S!CNOD'38.
(Closed·pattern) N. Pasquier, Y. Bastide, R. Taouil, and L. Lakhal.
Discovering frequent closed itemsets for association rules. !CDT'33.
(Sequential pattern) R. Agrawal and R. Srikant. Nining sequential
patterns. !CDE'3S
October 7, 2011 Data Nining: Concepts and Techniques 83
Ref: Apriori and !ts !mprovements
R. Agrawal and R. Srikant. Fast algorithms for mining association rules.
vLDB'34.
H. Nannila, H. Toivonen, and A. !. verkamo. Efficient algorithms for
discovering association rules. KDD'34.
A. Savasere, E. Omiecinski, and S. Navathe. An efficient algorithm for
mining association rules in large databases. vLDB'3S.
]. S. Park, N. S. Chen, and P. S. Yu. An effective hash·based algorithm
for mining association rules. S!CNOD'3S.
H. Toivonen. Sampling large databases for association rules. vLDB'36.
S. Brin, R. Notwani, ]. D. Ullman, and S. Tsur. Dynamic itemset
counting and implication rules for market basket analysis. S!CNOD'37.
S. Sarawagi, S. Thomas, and R. Agrawal. !ntegrating association rule
mining with relational database systems: Alternatives and implications.
S!CNOD'38.
October 7, 2011 Data Nining: Concepts and Techniques 84
Ref: Depth·First, Projection·Based FP Nining
R. Agarwal, C. Aggarwal, and v. v. v. Prasad. A tree projection
algorithm for generation of frequent itemsets. ]. Parallel and
Distributed Computing:02.
]. Han, ]. Pei, and Y. Yin. Nining frequent patterns without candidate
generation. S!CNOD' 00.
]. Pei, ]. Han, and R. Nao. CLOSET: An Efficient Algorithm for Nining
Frequent Closed !temsets. DNKD'00.
]. Liu, Y. Pan, K. Wang, and ]. Han. Nining Frequent !tem Sets by
Opportunistic Projection. KDD'02.
]. Han, ]. Wang, Y. Lu, and P. Tzvetkov. Nining Top·K Frequent Closed
Patterns without Ninimum Support. !CDN'02.
]. Wang, ]. Han, and ]. Pei. CLOSET+: Searching for the Best
Strategies for Nining Frequent Closed !temsets. KDD'03.
C. Liu, H. Lu, W. Lou, ]. X. Yu. On Computing, Storing and Ouerying
Frequent Patterns. KDD'03.
October 7, 2011 Data Nining: Concepts and Techniques 8S
Ref: vertical Format and Row Enumeration Nethods
N. ]. Zaki, S. Parthasarathy, N. Ogihara, and W. Li. Parallel algorithm
for discovery of association rules. DAN!:37.
Zaki and Hsiao. CHARN: An Efficient Algorithm for Closed !temset
Nining, SDN'02.
C. Bucila, ]. Cehrke, D. Kifer, and W. White. DualNiner: A Dual·
Pruning Algorithm for !temsets with Constraints. KDD'02.
F. Pan, C. Cong, A. K. H. Tung, ]. Yang, and N. Zaki , CARPENTER:
Finding Closed Patterns in Long Biological Datasets. KDD'03.
October 7, 2011 Data Nining: Concepts and Techniques 86
Ref: Nining Nulti·Level and Ouantitative Rules
R. Srikant and R. Agrawal. Nining generalized association rules.
vLDB'3S.
]. Han and Y. Fu. Discovery of multiple·level association rules from
large databases. vLDB'3S.
R. Srikant and R. Agrawal. Nining quantitative association rules in
large relational tables. S!CNOD'36.
T. Fukuda, Y. Norimoto, S. Norishita, and T. Tokuyama. Data mining
using two·dimensional optimized association rules: Scheme,
algorithms, and visualization. S!CNOD'36.
K. Yoda, T. Fukuda, Y. Norimoto, S. Norishita, and T. Tokuyama.
Computing optimized rectilinear regions for association rules. KDD'37.
R.]. Niller and Y. Yang. Association rules over interval data.
S!CNOD'37.
Y. Aumann and Y. Lindell. A Statistical Theory for Ouantitative
Association Rules KDD'33.
October 7, 2011 Data Nining: Concepts and Techniques 87
Ref: Nining Correlations and !nteresting Rules
N. Klemettinen, H. Nannila, P. Ronkainen, H. Toivonen, and A. !.
verkamo. Finding interesting rules from large sets of discovered
association rules. C!KN'34.
S. Brin, R. Notwani, and C. Silverstein. Beyond market basket:
Ceneralizing association rules to correlations. S!CNOD'37.
C. Silverstein, S. Brin, R. Notwani, and ]. Ullman. Scalable
techniques for mining causal structures. vLDB'38.
P.·N. Tan, v. Kumar, and ]. Srivastava. Selecting the Right
!nterestingness Neasure for Association Patterns. KDD'02.
E. Omiecinski. Alternative !nterest Neasures for Nining
Associations. TKDE'03.
Y. K. Lee, W.Y. Kim, Y. D. Cai, and ]. Han. CoNine: Efficient Nining
of Correlated Patterns. !CDN'03.
October 7, 2011 Data Nining: Concepts and Techniques 88
Ref: Nining Other Kinds of Rules
R. Neo, C. Psaila, and S. Ceri. A new SOL·like operator for mining
association rules. vLDB'36.
B. Lent, A. Swami, and ]. Widom. Clustering association rules.
!CDE'37.
A. Savasere, E. Omiecinski, and S. Navathe. Nining for strong
negative associations in a large database of customer transactions.
!CDE'38.
D. Tsur, ]. D. Ullman, S. Abitboul, C. Clifton, R. Notwani, and S.
Nestorov. Ouery flocks: A generalization of association·rule mining.
S!CNOD'38.
F. Korn, A. Labrinidis, Y. Kotidis, and C. Faloutsos. Ratio rules: A new
paradigm for fast, quantifiable data mining. vLDB'38.
K. Wang, S. Zhou, ]. Han. Profit Nining: From Patterns to Actions.
EDBT'02.
October 7, 2011 Data Nining: Concepts and Techniques 83
Ref: Constraint·Based Pattern Nining
R. Srikant, O. vu, and R. Agrawal. Nining association rules with item
constraints. KDD'37.
R. Ng, L.v.S. Lakshmanan, ]. Han S A. Pang. Exploratory mining and
pruning optimizations of constrained association rules. S!CNOD'38.
N.N. Carofalakis, R. Rastogi, K. Shim: SP!R!T: Sequential Pattern
Nining with Regular Expression Constraints. vLDB'33.
C. Crahne, L. Lakshmanan, and X. Wang. Efficient mining of
constrained correlated sets. !CDE'00.
]. Pei, ]. Han, and L. v. S. Lakshmanan. Nining Frequent !temsets
with Convertible Constraints. !CDE'01.
]. Pei, ]. Han, and W. Wang, Nining Sequential Patterns with
Constraints in Large Databases, C!KN'02.
October 7, 2011 Data Nining: Concepts and Techniques 30
Ref: Nining Sequential and Structured Patterns
R. Srikant and R. Agrawal. Nining sequential patterns: Ceneralizations
and performance improvements. EDBT'36.
H. Nannila, H Toivonen, and A. !. verkamo. Discovery of frequent
episodes in event sequences. DAN!:37.
N. Zaki. SPADE: An Efficient Algorithm for Nining Frequent Sequences.
Nachine Learning:01.
]. Pei, ]. Han, H. Pinto, O. Chen, U. Dayal, and N.·C. Hsu. PrefixSpan:
Nining Sequential Patterns Efficiently by Prefix·Projected Pattern
Crowth. !CDE'01.
N. Kuramochi and C. Karypis. Frequent Subgraph Discovery. !CDN'01.
X. Yan, ]. Han, and R. Afshar. CloSpan: Nining Closed Sequential
Patterns in Large Datasets. SDN'03.
X. Yan and ]. Han. CloseCraph: Nining Closed Frequent Craph
Patterns. KDD'03.
October 7, 2011 Data Nining: Concepts and Techniques 31
Ref: Nining Spatial, Nultimedia, and Web Data
K. Koperski and ]. Han, Discovery of Spatial Association Rules in
Ceographic !nformation Databases, SSD'3S.
O. R. Zaiane, N. Xin, ]. Han, Discovering Web Access Patterns and
Trends by Applying OLAP and Data Nining Technology on Web Logs.
O. R. Zaiane, ]. Han, and H. Zhu, Nining Recurrent !tems in
Nultimedia with Progressive Resolution Refinement. !CDE'00.
D. Cunopulos and !. Tsoukatos. Efficient Nining of Spatiotemporal
Patterns. SSTD'01.
October 7, 2011 Data Nining: Concepts and Techniques 32
Ref: Nining Frequent Patterns in Time·Series Data
B. Ozden, S. Ramaswamy, and A. Silberschatz. Cyclic association rules.
!CDE'38.
]. Han, C. Dong and Y. Yin, Efficient Nining of Partial Periodic Patterns
in Time Series Database, !CDE'33.
H. Lu, L. Feng, and ]. Han. Beyond !ntra·Transaction Association
Analysis: Nining Nulti·Dimensional !nter·Transaction Association Rules.
TO!S:00.
B.·K. Yi, N. Sidiropoulos, T. ]ohnson, H. v. ]agadish, C. Faloutsos, and
A. Biliris. Online Data Nining for Co·Evolving Time Sequences. !CDE'00.
W. Wang, ]. Yang, R. Nuntz. TAR: Temporal Association Rules on
Evolving Numerical Attributes. !CDE'01.
]. Yang, W. Wang, P. S. Yu. Nining Asynchronous Periodic Patterns in
Time Series Data. TKDE'03.
October 7, 2011 Data Nining: Concepts and Techniques 33
Ref: !ceberg Cube and Cube Computation
S. Agarwal, R. Agrawal, P. N. Deshpande, A. Cupta, ]. F. Naughton,
R. Ramakrishnan, and S. Sarawagi. On the computation of
multidimensional aggregates. vLDB'36.
Y. Zhao, P. N. Deshpande, and ]. F. Naughton. An array·based
algorithm for simultaneous multidi·mensional aggregates.
S!CNOD'37.
]. Cray, et al. Data cube: A relational aggregation operator
generalizing group·by, cross·tab and sub·totals. DAN!: 37.
N. Fang, N. Shivakumar, H. Carcia·Nolina, R. Notwani, and ]. D.
Ullman. Computing iceberg queries efficiently. vLDB'38.
S. Sarawagi, R. Agrawal, and N. Negiddo. Discovery·driven
exploration of OLAP data cubes. EDBT'38.
K. Beyer and R. Ramakrishnan. Bottom·up computation of sparse
and iceberg cubes. S!CNOD'33.
October 7, 2011 Data Nining: Concepts and Techniques 34
Ref: !ceberg Cube and Cube Exploration
]. Han, ]. Pei, C. Dong, and K. Wang, Computing !ceberg Data
Cubes with Complex Neasures. S!CNOD' 01.
W. Wang, H. Lu, ]. Feng, and ]. X. Yu. Condensed Cube: An
Effective Approach to Reducing Data Cube Size. !CDE'02.
C. Dong, ]. Han, ]. Lam, ]. Pei, and K. Wang. Nining Nulti·
Dimensional Constrained Cradients in Data Cubes. vLDB'01.
T. !mielinski, L. Khachiyan, and A. Abdulghani. Cubegrades:
Ceneralizing association rules. DAN!:02.
L. v. S. Lakshmanan, ]. Pei, and ]. Han. Ouotient Cube: How to
Summarize the Semantics of a Data Cube. vLDB'02.
D. Xin, ]. Han, X. Li, B. W. Wah. Star·Cubing: Computing !ceberg
Cubes by Top·Down and Bottom·Up !ntegration. vLDB'03.
October 7, 2011 Data Nining: Concepts and Techniques 3S
Ref: FP for Classification and Clustering
C. Dong and ]. Li. Efficient mining of emerging patterns:
Discovering trends and differences. KDD'33.
B. Liu, W. Hsu, Y. Na. !ntegrating Classification and Association
Rule Nining. KDD'38.
W. Li, ]. Han, and ]. Pei. CNAR: Accurate and Efficient
Classification Based on Nultiple Class·Association Rules. !CDN'01.
H. Wang, W. Wang, ]. Yang, and P.S. Yu. Clustering by pattern
similarity in large data sets. S!CNOD' 02.
]. Yang and W. Wang. CLUSEO: efficient and effective sequence
clustering. !CDE'03.
B. Fung, K. Wang, and N. Ester. Large Hierarchical Document
Clustering Using Frequent !temset. SDN'03.
X. Yin and ]. Han. CPAR: Classification based on Predictive
Association Rules. SDN'03.
October 7, 2011 Data Nining: Concepts and Techniques 36
Ref: Stream and Privacy·Preserving FP Nining
A. Evfimievski, R. Srikant, R. Agrawal, ]. Cehrke. Privacy Preserving
Nining of Association Rules. KDD'02.
]. vaidya and C. Clifton. Privacy Preserving Association Rule Nining
in vertically Partitioned Data. KDD'02.
C. Nanku and R. Notwani. Approximate Frequency Counts over
Data Streams. vLDB'02.
Y. Chen, C. Dong, ]. Han, B. W. Wah, and ]. Wang. Nulti·
Dimensional Regression Analysis of Time·Series Data Streams.
vLDB'02.
C. Ciannella, ]. Han, ]. Pei, X. Yan and P. S. Yu. Nining Frequent
Patterns in Data Streams at Nultiple Time Cranularities, Next
Ceneration Data Nining:03.
A. Evfimievski, ]. Cehrke, and R. Srikant. Limiting Privacy Breaches
in Privacy Preserving Data Nining. PODS'03.
October 7, 2011 Data Nining: Concepts and Techniques 37
Ref: Other Freq. Pattern Nining Applications
Y. Huhtala, ]. Karkkainen, P. Porkka, H. Toivonen. Efficient
Discovery of Functional and Approximate Dependencies Using
Partitions. !CDE'38.
Pattern Extraction with Fascicles. vLDB'33.
T. Dasu, T. ]ohnson, S. Nuthukrishnan, and v. Shkapenyuk.
Nining Database Structure, or How to Build a Data Ouality
Browser. S!CNOD'02.

.47.5  0: 0 1?.5 
..5.51 035.4/9 339;:9:91
"0;6/9  .;. 556507;:.51'058: 

.7;9 55985;#.;;95:  ::60.;65.516993.;65: 
.:006507;:.51.96.14.7 05;.51:0.3./3985;;4:;455 4;61: 55.96:251:6.::60.;6593: 964.::60.;65455;606993.;65 .5.3:: 65:;9.5;

::60.9 "0.65455 &44.6/9  . 556507.:1../.:...51'058:  .

:96.95 ..3 435:2 .79610.79: ..:.51..955..251:6!..9..1...#..51&...::60. 6985.64..790...:.6.9 9.93.6.9/5.6..0 .::/16045.: 7730.4: :/:850: :/:.5790.3:: 985...9:5:.6593455 6.: .65515595...1..:1..3303...9:  .0.6009:985..7.:519 . ..::.5.:985.35. 9:.3:: 096::.:/:85..95.:2.90.65: .065.5...4:.79676:1/9..# .511..:51.4(&*5.7..9..

5 0.5.4 ..5.51'058:  .6/9  . 556507.92.3:: "0.5 .3:: /36 0302:.47.3:: .5.51!:850...:.4.9.361:5 :.30..

5.:.3 :..9.5.510..:98 #.....65 .:....: 694:.514769.. :036::..5...90.:2: ::60..65 06993.95:0.65694.61.31.3:: &85.79679.3.651.455.5::5.55.3   :/.95 554769..

95: #. .6.5.3::5:7..41..4.7 7.95.4769.3 43....9.

65..::60.03. :9: .41.95.9.::0.51:..9.5. 3....7.3::985...65 3:.::0.

./..510/.:103:..95 .96:50/90/.

7730.1.51'058:  .. &4. 556507..:..01.9.6/9  .:003: 96.65: "0...06479::65.15.5.

.:985.95:.:06507..65.65%3: '9..0.5:.51 ::60..#.

/3..5:.:745  06545  98 #.<  2> 51.6593:         :894207 -:8-007 "0.6/9  .33.65.9.5. .5:...3 796/.5: .1      :894207 -:8-49 ...5:  065150 0 0651.507 .4:.                  :894207 -:8/.65065.9.0.51'058:  ... <    > ::60./3.4544 :7769..65 .... 556507.0.93:  ..51065150 :7769.3:6065. ..:. . : 796/.4:/6..

..51 .36:1#..95:.

95: 3657..95065..5:...69.354/96:/.064/5.#...

.95:   <.>065.5:               :/..  .. 7.

65 5036:17.95:     &63.514.....7.95: ..

7.:56 :79. :036:1:985..9:..1 5.95: 5:.4:..51. .

.:.:89 .4:.4. ..95   .7.:.4:7769.3 '  5...: 79676:1/#..

9 :.95 :985.51.:56985....:79.7.

67..36::3::06479::65698 7..7..5193: "0.95: %105....51'058:  .. 556507.95:..:.95:.916& " 36:17..95  79676:1/ ..6/9  ..

..36:1#..51 .95:.

 .:.95: 90: <..  ..#. . .  .:7 .:..> 5.:.64.. ..6036:1.  .:.  .4:.

6/9  .:...7.:.95:  "0.  .6.. 556507..337. ...:.95 .51'058:  ..

7 05.606993.96.65 ..65: .516993.5.65.5...6593: 964.3.61: 55...7.455 4..51:0.::60..::60.:006507.95:  ::60.14.9.51.#.65455.9 55985...96:251:6./3985.:.3:: 65:..4:.

51'058:  .65455 &44... 556507.:1.9 "0./.::60.6/9  ..:.

61:'94.0: 7969 9.3.65./34554.694.79 5..5:</9 1.3:6 065.2 :.. </9 1...5 # 5 & " 9. 98 7..0 .:>.7796.91036:9 79679.95: '165.69.79 5..4:..3.6985.7796.79>   9..:> :985. #96.0.6/9  ..&0.4:.985..94..:....5</9 1.#.7.9.51'058:  ./985.0.6 &  "0.5. 556507..3 &92.79> &0.6.9596. :6:</9  1.95: 5:/:..5:.61:69 55985.31./3 .

..7969.65.511.59.

.51.

985. .4:.6.553.:. .:63156..33 :0...5650.::79:.3  .3 &92./59.1 9.  .':..1.9:.5.0 79697955795073.7796.0: 5985..61 5.5 .

0..51'058:  ..556985.1 "0. .0.. 2  0.2 985..511.. 35.:. .511.690.5/ 59. '945.: ':.. 556507.:96435.:.6/9  .:. 59..4:....511.4:..5:.4:.

<  > 7/ 8.-.'7969369.473 .4:             &745   89 8..3  .4:. <  > :7   "0. < > < > < > < > < > < >  3/ 8.9.4:.4:.4:.4:.3 .4:..80% '1     .3 . < > < > < > < > < > < >  .45.:. 556507.. :7      <> <> <> <>   .51'058: . < > < > < > < > :7     .4:..6/9  .. <> <> <> <> <> :7      :7       .

4 #:16.'7969369.

 556507.6. 51..:52  .:7769.65.511.9065.065.../.4:.511..6:2  <985.511.5:..45.0.6:2 2 985.4:> 69 2 2 X2 16/5 2  0..:59.......19642 69.:52  .515.330.061 2.9... 51 9.0 ..4:.. 2  0.51'058:  .:16 50945..511.95 2 2 "0.:.6/9  .

..: &.659..511.0..7:3..3:67969 6.5.4769.

.77955 6.6552 &..:7769.6065..511.: .511.:60.4736.

65 <./1 .01 .59..0 /01> &3./0 .

/01964.1 :56.:.655  .01 :9461/0.51..0 #955 . 556507./01> "0../0 .51'058:  ./1 .01 964.:.51.6/9  .01 .5 <.

6;659.;.511.;:
&776:;;4:52

 .93:;15.56919 &;7:3

6552

4 7 .42.6 2 :30.4 Ã 7 .5.7 . 5:9.

42. 8 .

 9642.

 7 2.

48 .42.4 Ã 7 .8 97 .

8 .42.

42. 7 .

 8 .42.

 &.33.:052 16 69.33 2.4:.77955 69.

 .

:/:.::6016  ::56.52.

513. .0 9642 "0..6/9  ..:. 556507.51'058:  .

:60.:..511.:.:6.511.5/9 "5.511.654.:0.6.065.665..4:.6...9:.6915..796/34 '.511..:.354/960.54.9.0.5:7769..: 065..5:.511.: ..50.61 .&7769.

.9 .:.5616.

65 "0.511. 556507.:065.:.6551:.3:.6.9.6/9  ..0....51 065.33.:./3 &/:.51'058:  ..4:.0.5:.50.515 ...969561 065.:..5:.: 5.9065..5:..

47365.: \$:-8091:3.:6.6/9  ..:.5&7769.51'058:  ...511.943             "0.38. 556507.943             %7...

65679695&\$ .91.506.47345.05.66179694.679&\$ &\$...6.

7796.  /.365 .0:.:1.2:66/0.

...5:65:32:  ": '.::60.:.. 556507.1479645.65: 5& " "0..: ::.9.65.6593455.3 5.:.6/9  .: ./350.31.95...514730..65.9.. & '64./.6919:64.4:3.65:.5 ..3.51'058:  .5...0 . & &.51% 9.93.93.

:::6.6/9  .9.: ....16:7769./.5:6.560.5: &95254/960..31.: 47965796959.95 55 .335: 3.5690.::0..:7769.: "0..0.: 54/960.51'058:  .065.9. 556507.065.03......511.0.: %107.511...511./..651.73:0.5:..#.511.651.: '16:69236.:.335:6985.5:.

95:  &. 556507..9.36/...:"53'0 5.:.95: &0./.:9  "405:2 ....5065:631..9.. 505.5..65&0..54:..65:6 &0.:./.656.91.651.. 7...469455.#.9..7.4:. ..5151360...33985..369./.:.3985.3...51& !.6553.::60.3985./ 985.....:76.5.51'058:  .7.6/9  .57.5.:: 5  "0.

#%10.: 2.511..!4/96.

6:0699:76515. / 0 1  .:5.556...1 .:5/02.></1 / 1> 985..4:.:./985.511.9:6310.065. .9:<./ ..: /36.

0. / 1  ../:56...511.4:...

..1 .51#  50.92  5 .:46065.:...6 <.>:/36:7769.4:./ .9:631  #.

.6/9  ./.6593: 5& " "0. 556507.369.:1 .::60..:..469455.51'058:  .

../ .51'058:  .16.. 7....473002.473:57969 &0..69985././.. 556507.:6515 :.6514::1985.51.: 45985.9 0021 .51.4:.&....0  .0 &0.6/9  .#.:650..5..4736695..5:.31..47353.::60.::69.7.:..95:.473569985..65 93: 5 "0.95:./.91././01 5:..:.473 653/6919: 6036:96985...:...7.95:  '6655 &.95: &30..

.%10!4/96&0.51..91.56/5: "50.065.5:                "50/6.3335. .9451 985.

38.6593:69 4.  065..51& ':9 5.40. .51'058:  ..56 /5: %7.:2.9451985.:6.5  334.4:.. 556507...5.:/:.92. 5 & " "0..9438 5747  9028098  9028098 < 902809.:.99.0 & 95%  6.9 1.065.1./.5  .6/9   9028098  9028  9028 ..514730.

6..35026985.

57.:60.3 553657./.906:..95 55 3..:                   .555.:36.95:51:4.731. 6:0.5: 6.5159.5:.....7...511.::0.:::6 :0.: '651985..511.4:..

.511.35020.. 6..

59.65..

.51.

. 556507..610.:.5.65 "0.6/9  .:.. ..511.59.51'058:  ..

45=.7.6.65 963657.5:..59..33.. 556507..985..65::5360..95 ./0 1:.#.360.95 "0.7.95:964:69.511.6/9  ..3 985./0 . .4: .. 985.3985./0=..9..../01:..65:..5./0:..51'058:  .95:.0.55985.:.

90.65:.#.

1  2 4< 1 .65.5650 51 985. .1   4 < 1 -< .  .: %      9028-4:9 47/070/ 1706:0399028 1 . ..9964.. 2 5 - 2   &0.../. . 1 . .0. . .  8 5< . 0  5 2 3< 1 .5< ./ 1  .   2  5  23*8:55479 < 1 .'9. /   2 5< 1 . 2 5< .-0 9021706:03. .0.2< . 2 5< 0... ./07%.5:.

...4:.4:5 98501:0515 6919 ..95 .985.47. - - 5  &69. :53.

3:.#.  &0..90.5.5  065:.

6/9  ..9 "0.

.3:.

0.

/.

4.

.:.7 ..51'058: . 556507.

:6.5.#.

9 6473..0.65 647..9./.0.99.4:598501:05156919..3600995 .956.95 455 !9/9.31.565985.065.965 .5.5:.5:: %10993.4:.695.: 56.91 !9/3. 561..5..6/:.5694.5.3657.7.90..2.9&..469 985..6569985..469323..5:: #9:906473.

31 696550.51.065.352:.

. 556507.60631/69 "0.:. 06479::659..6/9  ..51'058:  .

006915.51...6..95:..9.5/7..#.7.6:/:.: .:: 985..9./.6515.95:0.65#...

3:; 

3:;

0

/.

4.

...567  #..95:..50/.54/.56.51565.5::...95:065.557 #.95:.7 #.569/ 4 7 #.95 6473...

915150 "0.51'058:  ...6/9  . 556507.:.

5#964#...51#.95:.

19...0651.985.5..3.9../35../.4..#.65.: &..

.#.9:.9 '9.

47 0043. 1 .: < .65.   ... 1..80 1 1. ..-0 1 9021706:03.5:6941797. - ...99073-./ 1  . 556507.0.47.-  0. - 5 43/943.99073-.3526.6/9  2 . 1.336...  .9. 2 5 .95/. 1.9/63365. 2  5  2 - 5 "0.: 6..5..:..37..66947: 0651./07%.- 1.51'058: .808 902 ..0985.2 .43/ 5.

.3#.964651..95.65.

65.::.3#./.6651.

9: 69..07...95.

././.#.90.: 0043.45..69.: 65:..0.065.

43/943.:..- 1706:039 5.2  1 1..985..! 9700  ./07%. ..2 .2 ... - - 5 - 2 2 .0. 2 .2  .99073870.: < 1 ..9094 2 < 2 12 .43/943. 1.   2  5  ..80 1. 7.2 . 1. 2 5 "0.2 1. 556507.6/9  0.  ../ 1  .51'058: .99073-.969.5.4:6.-0 9021706:03.95/..

3#.%09:65 55.0651.65.

:.9 < < 1 .! 9700 "0. 2 .4 0 1 . .43/943.:60.43/943.95/..! 9700 < 651 7.! 9700 651 7.. .! 9700 651 7.6/9  .43/943..95/.2 .:604  < 1 ..:6..95/....51'058:  .4  1 .2 .. 556507.2 .43/943..

&70.3.:&53#9#.5#..

65.. 0651.9 &776:.3 #.

.9'.:.91 :5379.:.

: < 9 9 /4 2 "0.5.9.3 ..4559:3...656.:53797.:6..51'058: .3 .3 /4 2 2 2  .6 7..3 .67.3 .6/9   .: < ..# 550.5.5/106476:15.656.:.3 2 2 %10.665561 650..6.9. 556507.7.

#..#..55985.95:.

:0651....9: 1.9596... %09:396985./..9.3 7.:7..65.95..7..51 1.4 065:..985..90.65 .61 69.95.7...0985.95:/7..

3#.65./.: .51.:0651.5.

.7960::65.10651...3 #.9 %7.65..05309.

9:3.9 5..3.5#.

. 064/5...5:653 657..:537.3359.33.9:47..::/. 69...65:6.065.

7.7.: .. 985...51'058:  .060:.:.95 "0. 556507..6/9  ..

&0.35#.

/#960.96.65 #.

7.5145#.544697960.6.:5..:..90.556.65.65 9:.1: '5065:.1.67960...90.9.../.

.1 #.65 .3 "0.07960.9.969.9.058: #.3337960.51'058:  . 556507.6/9  ...65 : #.65::7..3337960.:.9.657960.006:.

9.#..65.

47 0. 61:2:7./4 / 0/7 0.9.36.:.657960./.6551:.:1#960.5  0.3337960.65 '9.0 #..47 #. 7.9.65:.

796 0.4 4.4 0/ 0.

/ 0. .796 0. 0.4.

796 0 0 0 /.

796 .

796  0 0/   04.

796    0.

796   .

.:.796    "0..6/9  .51'058: . 556507.

#.

&7769.:./3.6/9   ..96..: 7969&0.51'058:    "0. 556507. '9:631     #:3920 8 0.809%   ! 74 97:3920 57477:3920     \$:5547997084/  .3..        ..9.

#.

: '9.96.

'9:631 ..9.65&0.809%     #:3920 80.      "0. .51'058:     .3.6/9  ! 749 %700!740.#960.&7769. 556507..943   \$:5547997084/  .:./3..

:#.

96.559 1..

51..

.65 560..:: ".1:.3391.906:4.985. 06479::11..59.51:6.:2.06589 106476:/6.455.51.7.6 .:#.9 3.511.9......95:6/.006915.:./.69: 560...660:1:./..0..511..

/.4:.398.9:.9 5697.565.5360.1:0.:067:065.51/315:/ #.....: /.90.91.

.90..51'058:  ...6/9  . 556507.514..95:..05 "0.:.9 567.

.4730..65:6.. .514.:.61636 55036:1985.4:.

.95: "&'   55:85..5  #9&7.5..9.95: 9&7...5  65:.7.37.

/.95: 659..9..50/91./3065:...:14556985.5.:9: .7.064734.0/:.:   647.

.51.9.

0/5..51'058:  .369.6/9  .:. 556507..4 & " "0.

.. 59 55 .

3 4.5..7. :0...551985..95: :..4:      51 :0.69              '1    .551:7769.4:              #6.

7..95: &50:.4...

.9:0.....:.51'058:  .95:964 1.916 05. 556507.7...5 % .95 5651.:: 5& " "0...34553657./..6/9  .6002   53.

6..1..3:...55985.36:1#.:05156919 3:.33985.4:5:7769.95:"&' 3:.

.

.

0 #.3:6.4:..0.4: .5 %  .6 "&'505.51 #.0 51985.:7 '      .56.:    "0.469 55985.95:.6/9  .369. 556507..95  #  .:.36:1..51/.95:.51'058:  .036:17.1:. 0 1   ..9.51...5:.65.. 0 1  0   9.0 1:..95909:3 5. 0. 985.90:7. /  0   .. ..036:17.:0.

."&'  5536:1.95.:/ #.4:..

 &/.96.5 :491.495.77.4:. .9:5960099506 .

336 :1:051.7955  .:5.90.549.51:7  :7  .4:..65.5..65 6..97960.51.5/7951 /91.64.:.

77:0.3.9.

7960.65 '67.

1657:16.9.

 556507.:/:.7960..3985..:.0025 "0.5 :9.33: 650.4:2775./3.6/9  ../3:.964 .19.65 .933: 05.195..4:7769.:.51'058:  .5795.4.3.19..:.360....

%  55/736959.694. 9..3..1.3694.0...0..  <' ' > .

5: .6.3:.3:.9.

:.51'058:  .9..   <'> 03.0039.1:065.4:.:.55.3.5.6.2 :.03.2.:.3.0.   .65. . .51.3 & " %  . 556507.35.  .6.9:0..5.5:.  <' '> :.  .775.455 "5327.3  #% # &56.1: .:1659. .. 95036:17....  <' ' '> .6/9  .9.95:/..0261950:6..  .6&  "0.9 .: :51:.65: . ..0.

9..:6 55 .9479645.61: "#'  .3  7:.

:. 7.5 ....9.975..9 #...616945506515:1985.3  51.95 # .96.3396:/.4.:4.:.9 .90.5496:06345: 65:.

65.549.455 "0.:.51'058:  . 556507.96905...6/9  ..

7 "0...3..:.656::60..51'058:  .65%3:#3.59.6/9  .:. 556507.

..7 "0.65%3:%39.6/9  .:. 556507.:.3.656::60..51'058:  ..

65%3: & 5&.. 556507.51'058:  ..6/9  ..656::60.  "0.3.:.:..

/3985.51.:.65...3.95:  ::60..65 .9 55985.7.9.6593: 964..#.65455.4:..516993.96.14.455 4.61: 55..3:: 65:.51:0.65: .7 05.606993.:006507....::60.5.96:251:6.5.::60..

51'058:  .. 556507./...6/9  .65455 &44.::60.:.:1.9 "0.

..::60.506993.96:51:6::60...55.65 4543.::60. 556507.33.95: "0..9:.5.657.145:65.65 555.6/9  .::60..:......65 558.3..65%3: 5543..51'058:  .

73.55 3.

..1.3::60.369 :7769.5: .970.91 43..90: 3/3:7769.4:6.4:.. 7369..656:.56949.:.36933.65%3: ..6.

 8:55479 (  8:55479( \$2 8:55479( 0. 556507..3  &92.0 23*8:5 .33455 9..0 23*8:5 9101:7769.0 23*8:5 0.51'058:  .:. . 0..0 23*8:5 "0.6/9  0.5  5694:7769.5.

.3.

 556507.4: .6.93:./9151..65%151.1(:7769./9.473 432 . 065150 * .. 065150 *  432 :.5.3 /.9:.:.33::60.70.95 &6493:4.::7769.6.50:.:065193 93:9151..1 .93:.5...69 "0.51'058:  ..1(:7769.5.5.6/9  ..:165./9.696.1.50:.69 93.503.:036:.50:..65:7:/.

.55 3.

65 &53.3::60.45:65..

145:65.393: /:  432  /:  /9..1 3.

: 5.393:K 145:65:697910.145:65.9..

: ..145:65.  .17910.::60 93: 5697..

.65  :.15.  .  6007.  .

  /:  7670695  /:  062 /91.

.15.0: "0..145:65..0/.17910.6/9  .519..: /:  062 .65 03:.:..3:1:09..:5490 4730..3: 56 69195..9/.7796.3.3:1..465 ..69195.0 \$.54/9676::/3..690.51'058:  . 556507..95 .9/.465.7796...5.::60 93: 97.:5...

..65 8.9/.691/65490.90: 1. 9...50..5/0.01:09.93:   9...& "   3:.1  &.0/4..5..:1657915106507..65/..:.9 ......:1651.5.401:09..9....61:  5.: :0..5.3.95:.3 .9.69 :.55\$.1:..3 &92..::60.65: '058:0.9/.65/.

.5.55.::60.65   .65 :0.:./. 556507.515133 &4.303:..51'058:  .334.3.5  339& "  65145:65.95.:1.6/9  ..:4.65  .4.::60.5 9 69.5  "0...

.5: 593.&.01/9../.2..17969.9/..: 515.90 !490.: :09..3:....0:09..33985.5.9973..9.31..65.656\$..6455:506507.

: 339892 692 .0/:33:..7910./3:0.55..:..169455 '033:6.5: ..

.420 -:8 .:.0 3.9 "0. 556507.5/40.0/: 0.145:65.3 0/610699:7651..:.51'058: ..6/9  .420 -:8  . 7910.0 3.420 -:8 .420 ..6.0 -:8 3.:.0 3.: 559641..

.:...::60..93: 451:4..51164 !490.33 1:09..41 .915.9/.40.5::6.0.\$.4..65%3: #9676:1/5...1 &0.5.065150690647. &.

3:. .5 0..::60.05..5.6593: .3 93::5...6593:8..1...8..9.5 8.::60.669459.

.91 .473 .. 556507.420      -:8  7084:943%' "0.6/9  .0     3.:.51'058:  .

4651 4.3.065:..60099.93/.7730.65.5.5#.9 .641:745:700.9:..: ..95....55".65 '67./3 :.51.3  &64.9.4:   1.95: 3/3:7769.5.

91.3   .5 .7...6:70:745 /..67..95: .036:1985.

35.2 .45:469 1:9./3 5.339.::745 5#.40.

51:30.6/9  ..:.46:.9065:. 556507.7964:57.65....51 455 ..51'058:  .645 "0.90.

:006507.95:  ::60..65.9.96...5..516993.65455..7...6593: 964.455 4.14.51.96:251:6.:.606993.5.9 55985./3985.::60..::60.65: .4:.7 05.51:0..#.65.61: 55.3:: 65:.3..

:1.:..65455 &44.::60...51'058:  . 556507.9 "0./.6/9  .

33    &4 96    !   19  ! !  9.55:: .3 (   *:469.3:   73.33 .:3.:9617515./.15.:2.3 !6..509./.6.15 '69.51065150 .5.06993.3 (   *:4:3..33 56./..33 !6.369:7769./..09.:2.15.:2.:96993..3 &4 063     . .3.09.33 6:./.009..:..9:.  ./.09. 73.:2..65: ./.

 19       .

  .

6/9  19     .  "0.

     .

  .

  .. 556507.51'058: .:..

5 4. .:9: '.15  60:.33.* 902 * 8:5       . 19  !   ! !  6 32 4 0 4 ?0 4 ?4?0        !66 &4 063 4 0     ?4 0     4?0     !6 32 ?4 0 ?4 ?0 ?4 3.956..431  8:5  2.        &4 96 0 ?0 % ..661. * .51065150..6979:5.06993.35.:.65 .9 &9.93.9:.65: &64...51/ 661 .55::4.: /432 (  *:4:3.55.:9:66993.649:/432 &7769..

. 556507.:.4  8:5   :3.6/9  .065         06         /      .51'058: ..0780   "0.

9 . 6614.65:53.5:.956..33.3: .:9:&631:1 3.0 .65.9.:9:69 06993..51/ .0.

.065 69 069500631/ 6614.:9: "405:2' 6.33.

51'058:  .369.91036:9 79679. 05. 165.3   :/ "0. 556507.:...4: 0...6/9  .51 06950.5/19169 455 .065 ..

3:: 65:.4:.455 4..9 55985.606993./3985.14.65: .65..95:  ::60.6593: 964..::60..96.5..61: 55.5.3..96:251:6.:006507.:.9.7 05.51.516993..#.65...51:0.65455..::60.7.

:1.:./..::60.65455 &44. 556507.9 "0.6/9  ...51'058:  .

9.5.65:.

/.:1 \$9.

.9. 69.3:95..70.55.1  55 515....6/451:5.0 65:.9.0.:.1.65646:3 59..0 '7./.3:.90.9.95:0631/.95:5.7..5.5/.7960:: :9190.:.5..455 893.455:631/..664.1....56.60:1 .33 .9..

7961: 065:.6/ 451 &:.467.5.455065:.657369::0065:.:69 05.5.5...: 65./.4.9.9.9..:1455 :93/3.

/..6/9  .:.:1455 "0.51'058:  . 556507..

65 .65 .::60..65:.0 . 55 5631..  :5 &\$..5..065:.7065:.::0.9.5. 03.5..9..9.:5.

:4.3: :4  5.33:.55::065:.51 0:..6965 790 /9.. 593.:7769.96593:45.6.9.3289: 5179610.9.6490.5.6 5 0  45:6533065:.9::631. 556507.6/9  .95 065:.7.50.69:50.9.5...065150K  "0. K  45.5.51'058:  .99:/:.:.95:.69 %3 697. :.9:..3: 790  .

9.65:.5.51 55: 65:.9.

.9.9.90 65:.5.51455: 065:.:1&.

9.5.9105:./.95: :.337.:655 6..5...41.90:7.9.0 515.909.:: 515 :64 6965 .:5065:..5:9 5065:.9..:1:.

:1:.9.5.905 65:./.

55..0:.6/9  .:7:5:30.:43./.51'058:  .33 65:..65. 556507.9736:67 .9:.:897960::5989:..95455:.517.9..:.9..59:..:.651.65:1735897960::5 "0.4 65:.90 .90796/34656.9:.51455: 897960::55 & .9. .7:5 : 9:..

5..

9..#:5 % 23*8:5 5.565:.5.650. 656.

650. :616:.::79:.9.5..56. 5. 065:.&63.4:.. :4 & #90 A  :.5.: .55.4656.

.4656.65 :4 & #90 K  :56..5.

473 9.4656. A :.5 & 796.65 .5..

0./ '     '9.65 .: &616:9:79:. 4656.65 . 0 1   0    . / 0 1     #96.4 .5:./63.4:..6.. / 0 1  / 0 1    .   .

  .

   .

. 556507.  "0..:.6/9  .51'058: .

4:.&:.6/9  .: :79:. 5.55.5 & 796./:.65 .56.5.9.6965:. :4 & #90 K  :4656.650.0..65 ..   . / 0 1     #96.5. K  .#:5 % 23*8:5 656.9. 0 1   0    .4 .4:.65 45 & #90 A :4656.. 556507.:.650.. :616:.:: . 065:.51'058: '     '9.5:.6.656.:: &616:9:79:./ "0. / 0 1  / 0 1    .473 9..

  .

   .

  .

/.:  .5:.9. . :4 & #90 K  :56.:/:.6.6..5:: &0050...5.:.65 ::0050.36625.5.4.:30.6 1....&:.5/ 1.4::..4:.&:.9.:0050.::065:.656.9.:5 :/.  :79.&0050.4: 45 & #90 A  ::0050.5:./3655.5.5. "7.:5.0.651.:165    & 065.5:: 065:..:165.:0050.5:: 5 .0.9451/...5:.9..

 556507.:./3 "0.065.57:...6/9  .51'058:  .

51'058: .9.3 < < < < <  902809 < "0...:.. 556507.80 %     9028     902809 8:5 <   <  \$.3  902809 8:5 <   .4 .473 ..6/9  \$.3 <  <  <   902809 8:5 < < < <      902809 8:5  902809 8:5 < < < <     < < < < < <        902809 < \$..-.'7969369.

 556507.3 <  <  <   902809 8:5 < < < <      902809 8:5  902809 8:5 < < < <     < < < < < <        902809 < \$.3 < < < < <  902809 < "0. .9.-.47969 65:...80 %     9028     902809 8:5 <   <  \$.5.5...6/9  \$..9.:.9.51'058: .!. 369. &4<& 790>  .3  902809 8:5 <  65:.

55.9.'65:.4#: ..517969369.

.5. &4<& 790>  .80 %     9028     902809 8:5 <   <  \$.9.3  902809 8:5 <  65:.3 < < < < <  902809 < "0.7 .6/9  \$.9.9.. 556507.6565:..4656.-.:.3 <  <  <   902809 8:5 < < < <      902809 8:5  902809 8:5 < < < <     < < < < < <        902809 < \$...51'058: .5.

517969369. 45<& 790>  .3  902809 8:5 <  65:. &0050.6/:1  902809 < "0.51'058: .3 < < < < < 56.5.6/9  \$...65:.7 .4#:.9.80 %     9028     902809 8:5 <   <  \$.3 <  <  <   902809 8:5 < < < <      902809 8:5  902809 8:5 < < < <     < < < < < <        902809 < \$..9.. 556507.:.9.-.'65:.9.5..3 ..441.

.5.6.659.5.5.9..: % 23*8:5 659.6065:.5'665:.:5.9.

K  "919.45. 4656.4: .4:5.65694656. & 796.65/796793 69195.3.

: &616:../ ./ .1:0515 6919 .4:..5.5./ 63.   1 /  0  ../064:.

:.5:.6/9  .. 556507.65 "0.65 .   . / 0 1     '9.4656.51'058: '     .4 . / 0 1  / 0 1    . 0 1   0    #96..0.

  .

   .

  .

  K :0659.&.5.5.9.: .9653659../365:./3.

  1 /   0  . .  K :0659..65 9 .4 .4:.:0515 6919%.3.31:0515 6919%.1 .65 9 .4..:79 :0.:.4656.63.065:.4:.  .5. .. :6 16:9../34656.9.5.: .

1 :.51 .::. 0.96530659.4:. / 0 1     #96..51'058: .:.4:. 556507.: . . :6 16:./3 "0.5.6/9  .5.. 0  / 1    .9.:1 .79 ': .065:.   ..511.4 .  K ::.

  .

   .

  .

/3 56./365:. 0659..65569.5.5..4656.513659.9.57969.

:065:.5/4.556.57969455.5.369.6 . 7955/.9.4692 56190..0.33:9./7:117 5.4656..65 569:0050.9./7951 .:165.9.4 ..5.1 10.065:.1 .6985.0.0.1:.5.556.5..::4/3.5/7:15.163.065:.   &50.:: 796951:1.6.4:..

.9.. / 0 1     .4692 .4 .3   .7.95 96.

  .

   .

 "0.6/9  . 556507.51'058:  ...:.

4 .9.55.: .659.3      .5./365:.

 .

 .

31:0515 6919%./3..655.5:.0..   1 /  0  :0659.5.   45. .4:59.9.:7 :.

51.4656..4: .:.4:.65..5'650 9465985.65 9 .4:196771 .9661  #960. % &0.

5/0659.5./.6 .9.6065:.5.:0.5. .15.779679.5.691965.65 ..47960.:1455 476:5.

6/9  .65 "0.5:. 556507..51'058: .:.  1 / 0   1 / 0 .   1 /  0  % 23*8:5 '     '9.65 ..0.4656.  1 0     0   .

: 95.5.9..065:.989195.5.695 06530.9.7365:.5135 3.5.:4.4.

5.6 45985.9.6569196.9 0659.51 .:./. '5:5.5./3065:.9./. . 556507..11.51'058:  .:.9:.9.9:.9 :5606530.0699:76515 7960.6/9  .:65065:...5 . /6. % .60659..9:..:06530...69195 ./3 9 .6:.4: '9.:5.4:.6..5.: .56919% : .: "0.9065:.5.691969.

/3.5.5.3  A  :4 & K  ./3 65:.4:0631/6.6/9  659..4:0631/6.5. & A K  41.5.5.3  K  :4 & A  . .5 & A K  :4 & A  .9..5.4:0631/6..5.:9659.3  K  :4 & K  .4:0631/6.3  A   "0.9.65:.

:.65 &. 4656..65 659.9653 0659..51'058:  ./3 : : : !6 !6 : : : !6 : : !6 : : !6 !6 !6 !6 ./3 4656. 556507.

5.9.65:.

4656.5.Z & . Z & &  &  45 & A  45 & K  4.:.51'058:  .9 65:./3 56 : &0050.23 56 56 56 56 56 56 56 "0..65 56 56 : 56 : : 56 : 56 : 56 : 56 0659. & A  065.:1 5559. & K  065.6/9  .5 & K  . & A  4. 556507. & 7  7 Z < A K > :7769.K  9.Z & . & K  :4 & A  .23 . & A  5. & K  :7769../3 : 56 656.5 & A  9..9. : : : : : : : .65 : : 56 : 56 56 : 56 : 56 : 56 : 0659.3#0.K  :4 & K  .

::0./3 ..65 &.65665:.3. 659.9653 0659.9.5.4656.5../3 656.65 &0050.: 5.

:..65 .4656. 556507.6/9  659.65 50659./3 "0./3 4656..51'058:  .

7.51.::60.455 4...9 55985./3985.5.4:...96...5.:.95:  ::60.516993.51:0.606993.61: 55.65: .:006507..3.65..6593: 964.96:251:6.9.65.14.7 05.65455.3:: 65:.#..::60..

65455 &44...9 "0. 556507.:1.51'058:  ..::60.6/9  .:./.

985.

5./3985.65 .455 &0.95455.3..65.511..59..95 55&44..7.954554....54769.7.:.:251....#...9 985. #960..61: 7969 .

5...57.7796./..9. "&'  9.:1 #96.9:..515.95: 65:.693:..3694.0 %  55.9.0.

:.51'058:  .:1455 55:85.95: .917.65: "0./..7730.3..51..6/9  ..90.5:65:.. 556507..51:.

985.

.95 55%:.3.#..90#96/34: 55.

51:.65 55.: 5:9..90...65  4.3.65 13.985.3.336:34.. :85.657369..5..57.65   !:850..95:.3::..9:.95: #..91 7.95: &979:5 563 0650:  7730.935.1.639.51/6....5.

:..65 5:/31.6/9  .51'058:  .95 03..7.455 "0. 556507......::0.

:6..51 &.6593:/.3 ' 435:2 ./.65%3: % 9.91.%...::  & "  .::60..:06507.#.4:53..5:..4  55 ..:6985..95 55 ::60.

.95:964 1.95 %  ...7...:: & "   36:1..916 05../.34553657.

.::60.4:...7..95 ! #.5..63 .95 % 9.6/9  .036:1.95:   "0..:..:..3  :0695985.2.1 % '.51 .37.:69.:89  .51% &92....51'058:  ..  55:85.3.6593: '  &85. 556507.3 7.

6593:      .. 505./.5..  '6655 .51# &  50.4:69 1:0695...::    & #..4:69455.469 455.:.::60.:9  "405:2 ...6593:    &..51  92..:479645.553.6593:53.369.92  & 5 .369..: % 9.51.%7969.51% &92. .369.91.::60.::60.51& !.46 05....:.3.

93.::60.3 5.::60.65.95././.:::..5.514730.40./..5   334.6/9  ..31..6593:694. 065.:.:1.65:  & "  "0.51% 9...3:: & "  & &.4:3....92... & '64.5 .6593 455. 556507.4 69455.5...:2.9.6593: & "   '6655 &.369./...:.4:.::69.514730..47353..51'058:  .91.::60.51& ':9 5..5.9.6593:   & 95 %  6.: .

.%7.

9:.65. #960.

.65 & "   #  .97960.:1# 55 % ..95:..6 "&'505.51 :.4&.511.51   #9.333.369..46959.5  55985.3 ..9.5  # .5   .369.5  .65 .51%  .:  #.9/.656985.:       #.5 .51 5  55985.5  .4:.6.51 .0.1647.5:.26  55'67...469 55 985.5  .4:.51# '.:.9.5 .9.0#960.3  .36:1.:/ "7769.1 .7.. 59.65     .

51\$95 985.:.:         6    "5647.5 &.695.9.:.36:1 #.90569. &..95:   "0.985.95:.. 544&7769.36:1.51 # "&' &.5  .4:....    .6/9  .#.51'058:  ...:69 55985. 556507.6.5 .

 92  9 .::60.9.51  #. .4:.3694.:.9..369...3..6593:   .9. 55 &    03.  "..51%6549..65 .0.61:  .2 & #.51:.9.4 691:0696.3 59.2.6 % 505.333.%9.369.51 . .46936:1.

2 %#!'% 51536:1#. #955369.:.5 .:   #.3.:   "0.65:.469.9..4:...95:5656360.5  65    '5  ..:.51'058:  .5..6/9  . 556507..51 .:.

.% 55 3.

%3: % &92.3  5559..5.51\$..51% 9.3...31.5..::60.6593:     .51  :069643..73.5.

/.3.51' '62....../3: & "  ' 21.33...5.::   % &92.6593:5 3.3  558..5.::60.65...4.. .91.6.6 &  69:.455 :5.::60.. .51% 9..6593:964 3.   6946...993.

55.6/9  .   6946.369.9965:69...3..4: .35.  & "   4...6593:695.145:65.567.::60.51 5133 &. ::60.41..6 &  69:....  647.65%3:   "0.::60..4.51:.31.9.51'058:  ..4190..5 ::60.6593:   %   339.367.51 .. 556507.0.5..6593:&04  . ' 21.:.3'6969\$..51' '62.:.65 & "   61. .

.92.5 ..606993..90.3:..6593:..9:.:61:0691 .::60. 59.% 556993./3 .65:.35.9:.9:   # ...55  '6655 .65: & "   &39:.51 &39:./. # %652.3.5 .5 & 95 %  6.46 5155.5%3: 34.6593:   & 95 %  6.51 334.55   .:2.5 6514.5 &0.553.::60.:.058:694550.515.593:9643.51   92.9:.

 . ..5 6 505.9 .51'058:  .:9:69 55 ::60.95:    "405:2 3.9:..95:   "0.6/9  .55:: ... &30..5.95..5  4.:969::60. 5..:.:. 556507.%.1#.! '.9:..51 &9.5.65#.. 55 66993.65: '      4   ..51 ...

.951:6%3: %  6  #:.3.51& 9 5&\$.% 55".

..5:.6593:    5.:9  "405:2 ...65:5.65./.::60..649.6593:     &..65 %  6..../63  3.91.5 & /.32679..656.9.696 \$93602:59.3.51 164 3:.0.::60..6969455 .3..::60.  5569:.51&  !:.51& !.  &.:60:.4 .95..965 5.::60.5 .65:     ':9   334..

8.36..:6: %.1469.93455  & "   695  ./951:  6.693:5 7./31.455    .:. 55964#.1: ..9.60.:.95:.5 #96. 556507.5 & 6  .51 ..5...6/9  ..65:  ' "0..51'058:  ..

5.9.%65:.

::   "0.9.:.5.95:.51% 9...3.5.5 05.5 7369.5 ..6/9  .5106993.5  55&85.5.65:6065:.:   9.51 . 65:...::60.:   % !   & .4:.:1#./.95 55 % &92..6593:.%3.9..4 065:.: . 556507.5.69455.:5. \$  ..51 ..9.659.9.5  .2:4..5.:..::60.96.51'058:  .6  &4&#%'&85..2: % %.4556 065:..2:4.51.5...3  55.9.5 .95 55.51 795567.979::6565:..6593: & " ! .:    #  ./365:.9.5  55985.5  #.4.9.2:4.5.1:.3#.3#.5  ..:    #  .5....5 .51  & ..

3.5  #5.5.50479645.3.5179694....:850:   .51  92.05.6 \$ 5  .469 55985..65: .95: % &92.46 :0696985.&850:  .955  #  .91#.% 55&85.90...95:59.: '   .2 &#505.3  55:85.51&..3 . '6655 . 7:61:55.37..553...51 .51% 9.369.

..95:05.5 55&85.3#.. : #9&7.3/#9.

51 .5  .51'058:  .95:5..7:069    .95:   "0.51% :.1#.5.: &   .3 #.:.   9..5 .51 .7 #.5 36:9....9.#960.7 5536:1985.. 556507.9.6/9  ..9 36&7..97: 985.95 96.&/9.:..460..5 5536:1&85..

#969::%:63...41...6/9  .3::60.6.95: &&'  "0.5 :0696&7.4769.....51 '951:/7735"#.51 ':62..51   55%0995.    56736:.3 #.51/.5  5  ..51'058:  .. ..6: 05..41...65...5  .5 :0695/00::#.  679:2.:.. 556507./.95:.4:5 3..705694..3  3.51.65%3:5 69..% 55&7..5 . 556&7..51 .:: && " % .65%545. 55'0563665/6:    " % ..

.% 55985.95:5'4..#.

51 5 05.95: 5'4&9:.&9:.5  65. 556#.::60.51 .51 &3/9:0.6593:     ... 030.3#9610#...:..9.  "15 & %.5 6515./.:      5 ..4....9.4 .

'9.0.65 5.5:.3:: 55 3..65::60..

9.35.45:65.

65%3:  '"&  .'9.65::60.5:..0.

:6: .51  39: "535..  ! &1967636: ' 65:65   .36.. 55696.1:  ..

51'058:  .. '%'4769..635'4&850:    ...6/9  . 556507.:   .95:5 '4&9:.9/.65%3:65 635!490.5  .3.3::60. ' "0.5  .5 %  5...5 # &   55:509656:#9610#.:...

9..:    .51  7..51& &.6 #  :7.51/647.51  !.65  % %.65 5..9.3 #  :7....0647...3.99.9.5 ..51 .145:65..4.   !.29:5.656 43.3 % 9. "5.%0/9/.65 & ..

./.369.469:43.:1 .1.56:43.

65679...35967.. ..0/93.3.3.45:65.:  & "   9.3 ..9.65..69 59..9.

/ 096::.

./..51:/.

90.24.3:   ..9  .6.5 ! &..

%  6..9. 635.51!  116 :069.50/989:05. % 9..3 .3   & &.5 647..5 .51   334.

0/: '   9.656"#1.5 6.29:5...195 7369.4.51% %..64..

51'058:  .6/9  . 556507...656:7.510/90/: & "  "0.9: .:.70647...

6473 .4  # .51 .51 .5 647.5    5 ..51/7369.:9: & "  .7796.5  #  65 .50/9.6%105.5  55 3.0.../&     65  ... /:..5  .51   6515:1/5 0.65  .%0/9/.

5.9.2:4.&4.0:6.1: 59.51 /13.519.51 .5 /9.365:./:    ' 435:2  .6593:     & .9.5.9...6 &44.5 ..5     .5.5 \$6...15.::60.5  # ..35./     5  .:5.0. 45:65../6. &.

/5647.50/9 /:/'67.

516.65..64.

.:.65   "0..9.. 556507.51'058:  .6/9  .75.

.511950:      :   ..65..951:..::0.45564957.:165 3..733.9.51::60.53.95  65.65.65 %3 55     .%#693.5 ..51  05.::0...65.. 3. 5.::.513:.95: :0695.::0.5105.51 #  %009.

.51 .6/9  . 3:.:850 03:.95 :43.9 .51 :.9..36045.5.4:.95:5985.:.:.:165#910.5  .51'058:  .53..51# &  3:. 556507..65%3:    .95    5  .510..91.5 .900. &   5.99.::0.65/.. ::60.....5 #%3.::60.5 .5 &\$05.95/7.5  .65%3: &  "0.51 ..: & "   .

%&.4.51#9.0.9.

.. % 9.#9:95# 55  4:2 % &92.0.&.:69 .985065.651.65%3 55 59...51 3.52.1.5.     .5  3.0#9:95 556::60... .9.4:   5  65  ..9...0#9:95::60.5   .65 #9.3  92 #9.65%3:   ....51%  6.51 .5 77964.33#.

3::6'4. 45:65.3%9::655.

 556507.53.  .73'49. 59. 4.5. #..0: 5#9.....: !.51'058:  ...5.95:5.51# &   55985. 3.5533.5  #  ..6/9  .09..&9:.4:.&.. 55  4:2  92 .5#9.:...&..9.9. 55 #"&  "0.9.65.4:     ..51% &92.0#9:95.

  922 55 # #6922.0...06479::65.5 .:./.998 #..752  55..1.9.3.5.  '6655 05.6/9  .90.95.9.75150::5 #.3.631.51 &2. :069650.51% ! &4.29:5.9 .. 96:9 & "  "0.:003:   ' ...51 #. 556507..65:  .%".51'058:  .65..1:   ...65.9696..:&.: ' 65:65 &  .3...5177964..95 557730..65:     ..\$.