Module 3: Mining Frequent Patterns, Association and

Correlations: Basic Concepts and Methods

 Basic Concepts

 Frequent Itemset Mining Methods

 Which Patterns Are Interesting?—Pattern

Evaluation Methods

 Summary

1
 What Is 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, Imielinski, and Swami [AIS93] in the context
of frequent itemsets and association rule mining
 Motivation: 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.
2
 Why Is Freq. Pattern Mining Important?
 Freq. pattern: An intrinsic and important property of
datasets
 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: discriminative, frequent pattern analysis

 Cluster analysis: frequent pattern-based clustering

 Data warehousing: iceberg cube and cube-gradient

 Semantic data compression: fascicles

 Broad applications

3
 Basic Concepts: Frequent Patterns

Tid Items bought  itemset: A set of one or more

10 Beer, Nuts, Diaper items
20 Beer, Coffee, Diaper  k-itemset X = {x1, …, xk}
30 Beer, Diaper, Eggs  (absolute) support, or, support
40 Nuts, Eggs, Milk count of X: Frequency or
50 Nuts, Coffee, Diaper, Eggs, Milk occurrence of an itemset X
Customer Customer
 (relative) support, s, is the
buys both buys diaper fraction of transactions that
contains X (i.e., the probability
that a transaction contains X)
 An itemset X is frequent if X’s
support is no less than a minsup
Customer threshold
buys beer

4
 Basic Concepts: Association Rules
Tid Items bought  Find all the rules X  Y with
10 Beer, Nuts, Diaper
minimum support and confidence
20 Beer, Coffee, Diaper
30 Beer, Diaper, Eggs
 support, s, probability that a
40 Nuts, Eggs, Milk transaction contains X  Y
50 Nuts, Coffee, Diaper, Eggs, Milk
 confidence, c, conditional
Customer
buys both
Customer probability that a transaction
buys
having X also contains Y
diaper
Let minsup = 50%, minconf = 50%
Freq. Pat.: Beer:3, Nuts:3, Diaper:4, Eggs:3,
{Beer, Diaper}:3
Customer
buys beer  Association rules: (many more!)
 Beer  Diaper (60%, 100%)
 Diaper  Beer (60%, 75%)
5
 Closed Patterns and Max-Patterns
 A long pattern contains a combinatorial number of sub-
patterns, e.g., {a1, …, a100} contains (1001) + (1002) + … +
(110000) = 2100 – 1 = 1.27*1030 sub-patterns!
 Solution: Mine closed patterns and max-patterns instead
 An itemset X is closed if X is frequent and there exists no
super-pattern Y ‫ כ‬X, with the same support as X
(proposed by Pasquier, et al. @ ICDT’99)
 An itemset X is a max-pattern if X is frequent and there
exists no frequent super-pattern Y ‫ כ‬X (proposed by
Bayardo @ SIGMOD’98)
 Closed pattern is a lossless compression of freq. patterns
 Reducing the # of patterns and rules
6
 Closed Patterns and Max-Patterns
 Exercise. DB = {<a1, …, a100>, < a1, …, a50>}
 Min_sup = 1.
 What is the set of closed itemset?
 <a1, …, a100>: 1
 < a1, …, a50>: 2
 What is the set of max-pattern?
 <a1, …, a100>: 1
 What is the set of all patterns?
 !!
7
Computational Complexity of Frequent Itemset Mining

 How many itemsets are potentially to be generated in the worst case?

 The number of frequent itemsets to be generated is senstive to the
minsup threshold
 When minsup is low, there exist potentially an exponential number of
frequent itemsets
 The worst case: MN where M: # distinct items, and N: max length of
transactions
 The worst case complexty vs. the expected probability
 Ex. Suppose Walmart has 104 kinds of products
 The chance to pick up one product 10-4
 The chance to pick up a particular set of 10 products: ~10 -40
 What is the chance this particular set of 10 products to be frequent
103 times in 109 transactions?

8
Chapter 5: Mining Frequent Patterns, Association and
Correlations: Basic Concepts and Methods

 Basic Concepts

 Frequent Itemset Mining Methods

 Which Patterns Are Interesting?—Pattern

Evaluation Methods

 Summary

9
Scalable Frequent Itemset Mining Methods

 Apriori: A Candidate Generation-and-Test

Approach

 Improving the Efficiency of Apriori

 FPGrowth: A Frequent Pattern-Growth Approach

 ECLAT: Frequent Pattern Mining with Vertical Data

Format
10
 The Downward Closure Property and Scalable
Mining Methods
 The downward closure property of frequent patterns
 Any subset of a frequent itemset must be frequent

 If {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 & Srikant@VLDB’94)

 Freq. pattern growth (FPgrowth—Han, Pei & Yin

@SIGMOD’00)
 Vertical data format approach (Charm—Zaki & Hsiao

@SDM’02)
11
Apriori: A Candidate Generation & Test Approach

 Apriori pruning principle: If there is any itemset which is

infrequent, its superset should not be generated/tested!
(Agrawal & Srikant @VLDB’94, Mannila, et al. @ KDD’ 94)
 Method:
 Initially, scan DB once to get frequent 1-itemset
 Generate 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

12
 The Apriori Algorithm—An Example
Supmin = 2 Itemset sup
Itemset sup
Database TDB {A} 2
L1 {A} 2
Tid Items C1 {B} 3
{B} 3
10 A, C, D {C} 3
1st scan {C} 3
20 B, C, E {D} 1
{E} 3
30 A, B, C, E {E} 3
40 B, E
C2 Itemset sup C2 Itemset
{A, B} 1
L2 Itemset sup
{A, C} 2 2nd scan {A, B}
{A, C} 2 {A, C}
{A, E} 1
{B, C} 2 {A, E}
{B, C} 2
{B, E} 3
{B, E} 3 {B, C}
{C, E} 2
{C, E} 2 {B, E}
{C, E}

C3 Itemset
3rd scan L3 Itemset sup
{B, C, E} {B, C, E} 2
13
 The Apriori Algorithm (Pseudo-Code)
Ck: Candidate itemset of size k
Lk : frequent itemset of size k

L1 = {frequent items};
for (k = 1; Lk !=; k++) do begin
Ck+1 = candidates generated from Lk;
for each transaction t in database do
increment the count of all candidates in Ck+1 that are
contained in t
Lk+1 = candidates in Ck+1 with min_support
end
return k Lk; 14
 Implementation of Apriori

 How to generate candidates?

 Step 1: self-joining Lk
 Step 2: pruning
 Example of Candidate-generation
 L3={abc, abd, acd, ace, bcd}
 Self-joining: L3*L3
 abcd from abc and abd
 acde from acd and ace
 Pruning:
 acde is removed because ade is not in L3
 C4 = {abcd}
15
 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
 Method:
 Candidate itemsets are stored in a hash-tree
 Leaf node of hash-tree contains a list of itemsets and
counts
 Interior node contains a hash table
 Subset function: finds all the candidates contained in
a transaction

16
Counting Supports of Candidates Using Hash Tree

Subset function
Transaction: 1 2 3 5 6
3,6,9
1,4,7
2,5,8

1+2356

13+56 234
567
145 345 356 367
136 368
357
12+356
689
124
457 125 159
458

17
 Candidate Generation: An SQL Implementation
 SQL Implementation of candidate generation
 Suppose the items in Lk-1 are listed in an order
 Step 1: self-joining Lk-1
insert into Ck
select p.item1, p.item2, …, p.itemk-1, q.itemk-1
from Lk-1 p, Lk-1 q
where p.item1=q.item1, …, p.itemk-2=q.itemk-2, p.itemk-1 < q.itemk-
1
 Step 2: pruning
forall itemsets c in Ck do
forall (k-1)-subsets s of c do
if (s is not in Lk-1) then delete c from Ck
 Use object-relational extensions like UDFs, BLOBs, and Table functions for
efficient implementation [See: S. Sarawagi, S. Thomas, and R. Agrawal.
Integrating association rule mining with relational database systems:
Alternatives and implications. SIGMOD’98]
18
 Scalable Frequent Itemset Mining Methods

 Apriori: A Candidate Generation-and-Test Approach

 Improving the Efficiency of Apriori

 FPGrowth: A Frequent Pattern-Growth Approach

 ECLAT: Frequent Pattern Mining with Vertical Data Format

 Mining Close Frequent Patterns and Maxpatterns

19
 Further Improvement of the Apriori Method

 Major computational challenges

 Multiple scans of transaction database
 Huge number of candidates
 Tedious workload of support counting for candidates
 Improving Apriori: general ideas
 Reduce passes of transaction database scans
 Shrink number of candidates
 Facilitate support counting of candidates

20
 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, VLDB’95

DB1 + DB2 + + DBk = DB

sup1(i) < σDB1 sup2(i) < σDB2 supk(i) < σDBk sup(i) < σDB
 DHP: Reduce the Number of Candidates

 A k-itemset whose corresponding hashing bucket count is below the

threshold cannot be frequent count itemsets
 Candidates: a, b, c, d, e 35 {ab, ad, ae}
88 {bd, be, de}
 Hash entries
{ab, ad, ae}
.

.
. .
{bd, be, de}
.
 .

 … 102 {yz, qs, wt}

 Frequent 1-itemset: a, b, d, e Hash Table

 ab is not a candidate 2-itemset if the sum of count of {ab, ad, ae}

is below support threshold
 J. Park, M. Chen, and P. Yu. An effective hash-based algorithm for
mining association rules. SIGMOD’95
22
 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. In VLDB’96

23
 DIC: Reduce Number of Scans

ABCD
 Once both A and D are determined
frequent, the counting of AD begins
ABC ABD ACD BCD  Once all length-2 subsets of BCD are
determined frequent, the counting of BCD
begins
AB AC BC AD BD CD
Transactions
1-itemsets
A B C D
Apriori 2-itemsets
…
{}
Itemset lattice 1-itemsets
2-items
S. Brin R. Motwani, J. Ullman,
and S. Tsur. Dynamic itemset DIC 3-items
counting and implication rules for
market basket data. SIGMOD’97
24
 Scalable Frequent Itemset Mining Methods

 Apriori: A Candidate Generation-and-Test Approach

 Improving the Efficiency of Apriori

 FPGrowth: A Frequent Pattern-Growth Approach

 ECLAT: Frequent Pattern Mining with Vertical Data Format

 Mining Close Frequent Patterns and Maxpatterns

25
Pattern-Growth Approach: Mining Frequent Patterns
Without Candidate Generation
 Bottlenecks of the Apriori approach
 Breadth-first (i.e., level-wise) search
 Candidate generation and test
 Often generates a huge number of candidates
 The FPGrowth Approach (J. Han, J. Pei, and Y. Yin, SIGMOD’ 00)
 Depth-first search
 Avoid explicit candidate generation
 Major philosophy: Grow long patterns from short ones using local
frequent items only
 “abc” is a frequent pattern
 Get all transactions having “abc”, i.e., project DB on abc: DB|abc
 “d” is a local frequent item in DB|abc  abcd is a frequent pattern
26
 Construct FP-tree from a Transaction Database

TID items Items bought (ordered) frequent

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} min_support = 3
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} {}
Header Table
1. Scan DB once, find
frequent 1-itemset (single Item frequency head f:4 c:1
item pattern) f 4
c 4 c:3 b:1 b:1
2. Sort frequent items in a 3
frequency descending b 3 a:3 p:1
order, f-list m 3
p 3
3. Scan DB again, construct m:2 b:1
FP-tree
F-list = f-c-a-b-m-p p:2 m:1
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

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

{}
Header Table
f:4 c:1 Conditional pattern bases
Item frequency head
f 4 item cond. pattern base
c 4 c:3 b:1 b:1 c f:3
a 3
a fc:3
b 3 a:3 p:1
m 3 b fca:1, f:1, c:1
p 3 m:2 b:1 m fca:2, fcab:1
p fcam:2, cb:1
p:2 m:1
29
 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
Header Table
Item frequency head All frequent
f:4 c:1 patterns relate to m
f 4 {}
c 4 c:3 b:1 b:1 m,
 fm, cm, am,
a 3 f:3 
b 3 a:3 p:1 fcm, fam, cam,
m 3 c:3 fcam
p 3 m:2 b:1
p:2 m:1 a:3
m-conditional FP-tree
30
 Recursion: Mining Each Conditional FP-tree
{}

{} Cond. pattern base of “am”: (fc:3) f:3

c:3
f:3
am-conditional FP-tree
c:3 {}
Cond. pattern base of “cm”: (f:3)
a:3 f:3
m-conditional FP-tree
cm-conditional FP-tree

{}
Cond. pattern base of “cam”: (f:3) f:3
cam-conditional FP-tree

31
 A Special Case: Single Prefix Path in FP-tree

 Suppose a (conditional) FP-tree T has a shared

single prefix-path P
 Mining can be decomposed into two parts
{}  Reduction of the single prefix path into one node
a1:n1  Concatenation of the mining results of the two
a2:n2 parts
a3:n3
{} r1

b1:m1 C1:k1 a1:n1

 r1 =
a2:n2
+ b1:m1 C1:k1

C2:k2 C3:k3
a3:n3 C2:k2 C3:k3
32
 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
 Items in frequency descending order: the more
frequently occurring, the more likely to be shared
 Never be larger than the original database (not count
node-links and the count field)

33
 The Frequent Pattern Growth Mining Method
 Idea: Frequent pattern growth
 Recursively grow frequent patterns by pattern and

database partition
 Method
 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

34
 Scaling FP-growth by Database Projection
 What about if 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
 Project the DB in parallel for each frequent item
 Parallel projection is space costly
 All the partitions can be processed in parallel
 Partition projection
 Partition the DB based on the ordered frequent items
 Passing the unprocessed parts to the subsequent partitions

35
 Partition-Based Projection

 Parallel projection needs a lot Tran. DB

of disk space fcamp
fcabm
 Partition projection saves it fb
cbp
fcamp

p-proj DB m-proj DB b-proj DB a-proj DB c-proj DB f-proj DB

fcam fcab f fc f …
cb fca cb … …
fcam fca …

am-proj DB cm-proj DB
fc f …
fc f
fc f
36
 Performance of FPGrowth in Large Datasets

100
140
90 D1 FP-grow th runtime D2 FP-growth
80
D1 Apriori runtime 120 D2 TreeProjection
70 100
Run time(sec.)

Runtime (sec.)
60
80
50 Data set T25I20D10K Data set T25I20D100K
40 60

30 40
20
20
10
0 0
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2
Support threshold(%)
Support threshold (%)

FP-Growth vs. Apriori FP-Growth vs. Tree-Projection

37
 Advantages of the Pattern Growth Approach

 Divide-and-conquer:
 Decompose both the mining task and DB according to the
frequent patterns obtained so far
 Lead 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
 A good open-source implementation and refinement of FPGrowth
 FPGrowth+ (Grahne and J. Zhu, FIMI'03)

38
 Further Improvements of Mining Methods

 AFOPT (Liu, et al. @ KDD’03)

 A “push-right” method for mining condensed frequent pattern
(CFP) tree
 Carpenter (Pan, et al. @ KDD’03)
 Mine data sets with small rows but numerous columns
 Construct a row-enumeration tree for efficient mining
 FPgrowth+ (Grahne and Zhu, FIMI’03)
 Efficiently Using Prefix-Trees in Mining Frequent Itemsets, Proc.
ICDM'03 Int. Workshop on Frequent Itemset Mining
Implementations (FIMI'03), Melbourne, FL, Nov. 2003
 TD-Close (Liu, et al, SDM’06)

39
Extension of Pattern Growth Mining Methodology

 Mining closed frequent itemsets and max-patterns

 CLOSET (DMKD’00), FPclose, and FPMax (Grahne & Zhu, Fimi’03)

 Mining sequential patterns

 PrefixSpan (ICDE’01), CloSpan (SDM’03), BIDE (ICDE’04)

 Mining graph patterns

 gSpan (ICDM’02), CloseGraph (KDD’03)

 Constraint-based mining of frequent patterns

 Convertible constraints (ICDE’01), gPrune (PAKDD’03)

 Computing iceberg data cubes with complex measures

 H-tree, H-cubing, and Star-cubing (SIGMOD’01, VLDB’03)

 Pattern-growth-based Clustering
 MaPle (Pei, et al., ICDM’03)

 Pattern-Growth-Based Classification
 Mining frequent and discriminative patterns (Cheng, et al, ICDE’07)

40
 Scalable Frequent Itemset Mining Methods

 Apriori: A Candidate Generation-and-Test Approach

 Improving the Efficiency of Apriori

 FPGrowth: A Frequent Pattern-Growth Approach

 ECLAT: Frequent Pattern Mining with Vertical Data Format

 Mining Close Frequent Patterns and Maxpatterns

41
ECLAT: Mining by Exploring Vertical Data Format
 Vertical format: t(AB) = {T11, T25, …}
 tid-list: list of trans.-ids containing an itemset
 Deriving frequent 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) = {T1, T2, T3}, t(XY) = {T1, T3}
 Diffset (XY, X) = {T2}
 Eclat (Zaki et al. @KDD’97)
 Mining Closed patterns using vertical format: CHARM (Zaki &
Hsiao@SDM’02)
42
 Scalable Frequent Itemset Mining Methods

 Apriori: A Candidate Generation-and-Test Approach

 Improving the Efficiency of Apriori

 FPGrowth: A Frequent Pattern-Growth Approach

 ECLAT: Frequent Pattern Mining with Vertical Data Format

 Mining Close Frequent Patterns and Maxpatterns

43
 Mining Frequent Closed Patterns: CLOSET
 Flist: list of all frequent items in support ascending order
 Flist: d-a-f-e-c Min_sup=2
TID Items
 Divide search space 10 a, c, d, e, f
 Patterns having d 20 a, b, e
30 c, e, f
 Patterns having d but no a, etc. 40 a, c, d, f
50 c, e, f
 Find frequent closed pattern recursively
 Every transaction having d also has cfa  cfad is a
frequent closed pattern
 J. Pei, J. Han & R. Mao. “CLOSET: An Efficient Algorithm for
Mining Frequent Closed Itemsets", DMKD'00.
 CLOSET+: Mining Closed Itemsets by Pattern-Growth

 Itemset merging: if Y appears in every occurrence of X, then Y

is merged with X
 Sub-itemset pruning: if Y ‫ כ‬X, 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
 Item 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
 MaxMiner: Mining Max-Patterns
Tid Items
 1st scan: find frequent items
10 A, B, C, D, E
 A, B, C, D, E 20 B, C, D, E,
 2nd scan: find support for 30 A, C, D, F

 AB, AC, AD, AE, ABCDE

 BC, BD, BE, BCDE
Potential max-
 CD, CE, CDE, DE patterns
 Since BCDE is a max-pattern, no need to check BCD, BDE,
CDE in later scan
 R. Bayardo. Efficiently mining long patterns from
databases. SIGMOD’98
CHARM: Mining by Exploring Vertical Data Format
 Vertical format: t(AB) = {T11, T25, …}
 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) = {T1, T2, T3}, t(XY) = {T1, T3}
 Diffset (XY, X) = {T2}
 Eclat/MaxEclat (Zaki et al. @KDD’97), VIPER(P. Shenoy et
al.@SIGMOD’00), CHARM (Zaki & Hsiao@SDM’02)
Visualization of Association Rules: Plane Graph

48
Visualization of Association Rules: Rule Graph

49
Visualization of Association Rules
(SGI/MineSet 3.0)

50
Chapter 5: Mining Frequent Patterns, Association and
Correlations: Basic Concepts and Methods

 Basic Concepts

 Frequent Itemset Mining Methods

 Which Patterns Are Interesting?—Pattern

Evaluation Methods

 Summary

51
 Interestingness Measure: Correlations (Lift)
 play basketball  eat cereal [40%, 66.7%] is misleading
 The overall % of students eating cereal is 75% > 66.7%.
 play basketball  not eat cereal [20%, 33.3%] is more accurate,
although with lower support and confidence
 Measure of dependent/correlated events: lift

P( A B) Basketball Not basketball Sum (row)

lift  Cereal 2000 1750 3750
P( A) P( B)
Not cereal 1000 250 1250
2000 / 5000
lift ( B, C )   0.89 Sum(col.) 3000 2000 5000
3000 / 5000 * 3750 / 5000
1000 / 5000
lift ( B, C )   1.33
3000 / 5000 *1250 / 5000

52
 Are lift and 2 Good Measures of Correlation?

 “Buy walnuts  buy

milk [1%, 80%]” is
misleading if 85% of
customers buy milk
 Support and confidence
are not good to indicate
correlations
 Over 20 interestingness
measures have been
proposed (see Tan,
Kumar, Sritastava
@KDD’02)
 Which are good ones?

53
Null-Invariant Measures

54
 Comparison of Interestingness Measures
 Null-(transaction) invariance is crucial for correlation analysis
 Lift and 2 are not null-invariant
 5 null-invariant measures

Milk No Milk Sum (row)

Coffee m, c ~m, c c
No Coffee m, ~c ~m, ~c ~c
Sum(col.) m ~m 

Null-transactions Kulczynski
w.r.t. m and c measure (1927) Null-invariant

June 11, 2024 Data Mining: Concepts and Techniques Subtle: They disagree55
 Analysis of DBLP Coauthor Relationships
Recent DB conferences, removing balanced associations, low sup, etc.

Advisor-advisee relation: Kulc: high,

coherence: low, cosine: middle
 Tianyi Wu, Yuguo Chen and Jiawei Han, “
Association Mining in Large Databases: A Re-Examination of Its Measur
es
”, Proc. 2007 Int. Conf. Principles and Practice of Knowledge Discovery
in Databases (PKDD'07), Sept. 2007
56
 Which Null-Invariant Measure Is Better?
 IR (Imbalance Ratio): measure the imbalance of two
itemsets A and B in rule implications

 Kulczynski and Imbalance Ratio (IR) together present a

clear picture for all the three datasets D4 through D6
 D4 is balanced & neutral
 D5 is imbalanced & neutral
 D6 is very imbalanced & neutral
Chapter 5: Mining Frequent Patterns, Association and
Correlations: Basic Concepts and Methods

 Basic Concepts

 Frequent Itemset Mining Methods

 Which Patterns Are Interesting?—Pattern

Evaluation Methods

 Summary

58
 Research on Pattern Mining: A Road Map

59
 Chapter 7 : Advanced Frequent Pattern Mining
 Pattern Mining: A Road Map
 Pattern Mining in Multi-Level, Multi-Dimensional Space
 Mining Multi-Level Association
 Mining Multi-Dimensional Association
 Mining Quantitative Association Rules
 Mining Rare Patterns and Negative Patterns
 Constraint-Based Frequent Pattern Mining
 Mining High-Dimensional Data and Colossal Patterns
 Mining Compressed or Approximate Patterns
 Pattern Exploration and Application
 Summary
60
 Mining Multiple-Level Association Rules
 Items often form hierarchies
 Flexible support settings
 Items at the lower level are expected to have lower

support
 Exploration of shared multi-level mining (Agrawal &
Srikant@VLB’95, Han & Fu@VLDB’95)

uniform support reduced support

Level 1
Milk Level 1
min_sup = 5%
[support = 10%] min_sup = 5%

Level 2 2% Milk Skim Milk Level 2

min_sup = 5% [support = 6%] [support = 4%] min_sup = 3%

61
 Multi-level Association: Flexible Support and
Redundancy filtering
 Flexible min-support thresholds: Some items are more valuable but
less frequent
 Use non-uniform, group-based min-support
 E.g., {diamond, watch, camera}: 0.05%; {bread, milk}: 5%; …
 Redundancy Filtering: Some rules may be redundant due to
“ancestor” relationships between items
 milk  wheat bread [support = 8%, confidence = 70%]
 2% milk  wheat bread [support = 2%, confidence = 72%]
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

62
 Chapter 7 : Advanced Frequent Pattern Mining
 Pattern Mining: A Road Map
 Pattern Mining in Multi-Level, Multi-Dimensional Space
 Mining Multi-Level Association
 Mining Multi-Dimensional Association
 Mining Quantitative Association Rules
 Mining Rare Patterns and Negative Patterns
 Constraint-Based Frequent Pattern Mining
 Mining High-Dimensional Data and Colossal Patterns
 Mining Compressed or Approximate Patterns
 Pattern Exploration and Application
 Summary
63
 Mining Multi-Dimensional Association
 Single-dimensional rules:
buys(X, “milk”)  buys(X, “bread”)
 Multi-dimensional rules:  2 dimensions or predicates
 Inter-dimension assoc. rules (no repeated predicates)
age(X,”19-25”)  occupation(X,“student”)  buys(X, “coke”)
 hybrid-dimension assoc. rules (repeated predicates)
age(X,”19-25”)  buys(X, “popcorn”)  buys(X, “coke”)
 Categorical Attributes: finite number of possible values, no
ordering among values—data cube approach
 Quantitative Attributes: Numeric, implicit ordering among
values—discretization, clustering, and gradient approaches

64
 Chapter 7 : Advanced Frequent Pattern Mining
 Pattern Mining: A Road Map
 Pattern Mining in Multi-Level, Multi-Dimensional Space
 Mining Multi-Level Association
 Mining Multi-Dimensional Association
 Mining Quantitative Association Rules
 Mining Rare Patterns and Negative Patterns
 Constraint-Based Frequent Pattern Mining
 Mining High-Dimensional Data and Colossal Patterns
 Mining Compressed or Approximate Patterns
 Pattern Exploration and Application
 Summary
65
 Mining Quantitative 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 & Srikant@SIGMOD96)
3. Clustering: Distance-based association (e.g., Yang &
Miller@SIGMOD97)
 One dimensional clustering then association
4. Deviation: (such as Aumann and Lindell@KDD99)
Sex = female => Wage: mean=$7/hr (overall mean = $9)

66
 Static Discretization of Quantitative Attributes

 Discretized prior to mining using concept hierarchy.

 Numeric values are replaced by ranges
 In 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

(age) (income) (buys)
cuboid correspond to the
predicate sets
(age, income) (age,buys) (income,buys)
 Mining from data cubes
can be much faster
(age,income,buys)
67
Quantitative Association Rules Based on Statistical Inference
Theory [Aumann and Lindell@DMKD’03]
 Finding extraordinary and therefore interesting phenomena, e.g.,
(Sex = female) => Wage: mean=$7/hr (overall mean = $9)
 LHS: a subset of the population
 RHS: an extraordinary behavior of this subset
 The rule is accepted only if a statistical test (e.g., Z-test) confirms the
inference with high confidence
 Subrule: highlights the extraordinary behavior of a subset of the pop.
of the super rule
 E.g., (Sex = female) ^ (South = yes) => mean wage = $6.3/hr
 Two forms of rules
 Categorical => quantitative rules, or Quantitative => quantitative rules
 E.g., Education in [14-18] (yrs) => mean wage = $11.64/hr
 Open problem: Efficient methods for LHS containing two or more
quantitative attributes
68
 Chapter 7 : Advanced Frequent Pattern Mining
 Pattern Mining: A Road Map
 Pattern Mining in Multi-Level, Multi-Dimensional Space
 Mining Multi-Level Association
 Mining Multi-Dimensional Association
 Mining Quantitative Association Rules
 Mining Rare Patterns and Negative Patterns
 Constraint-Based Frequent Pattern Mining
 Mining High-Dimensional Data and Colossal Patterns
 Mining Compressed or Approximate Patterns
 Pattern Exploration and Application
 Summary
69
 Negative and Rare Patterns
 Rare patterns: Very low support but interesting
 E.g., buying Rolex watches
 Mining: Setting individual-based or special group-based
support threshold for valuable items
 Negative patterns
 Since it is unlikely that one buys Ford Expedition (an
SUV car) and Toyota Prius (a hybrid car) together, Ford
Expedition and Toyota Prius are likely negatively
correlated patterns
 Negatively correlated patterns that are infrequent tend to
be more interesting than those that are frequent
70
 Defining Negative Correlated Patterns (I)
 Definition 1 (support-based)
 If itemsets X and Y are both frequent but rarely occur together, i.e.,
sup(X U Y) < sup (X) * sup(Y)
 Then X and Y are negatively correlated
 Problem: A store sold two needle 100 packages A and B, only one
transaction containing both A and B.
 When there are in total 200 transactions, we have
s(A U B) = 0.005, s(A) * s(B) = 0.25, s(A U B) < s(A) * s(B)
 When there are 105 transactions, we have
s(A U B) = 1/105, s(A) * s(B) = 1/103 * 1/103, s(A U B) > s(A) * s(B)
 Where is the problem? —Null transactions, i.e., the support-based
definition is not null-invariant!

71
 Defining Negative Correlated Patterns (II)
 Definition 2 (negative itemset-based)
 X is a negative itemset if (1) X = Ā U B, where B is a set of positive
items, and Ā is a set of negative items, |Ā|≥ 1, and (2) s(X) ≥ μ
 Itemsets X is negatively correlated, if

 This definition suffers a similar null-invariant problem

 Definition 3 (Kulzynski measure-based) If itemsets X and Y are
frequent, but (P(X|Y) + P(Y|X))/2 < є, where є is a negative pattern
threshold, then X and Y are negatively correlated.
 Ex. For the same needle package problem, when no matter there are
200 or 105 transactions, if є = 0.01, we have
(P(A|B) + P(B|A))/2 = (0.01 + 0.01)/2 < є
72
 Chapter 7 : Advanced Frequent Pattern Mining

 Pattern Mining: A Road Map

 Pattern Mining in Multi-Level, Multi-Dimensional Space

 Constraint-Based Frequent Pattern Mining

 Mining High-Dimensional Data and Colossal Patterns

 Mining Compressed or Approximate Patterns

 Pattern Exploration and Application

 Summary
73
 Constraint-based (Query-Directed) Mining

 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
 Optimization: explores such constraints for efficient mining —
constraint-based mining: constraint-pushing, similar to push
selection first in DB query processing
 Note: still find all the answers satisfying constraints, not finding
some answers in “heuristic search”

74
 Constraints in Data Mining
 Knowledge type constraint:
 classification, association, etc.

 Data constraint — using SQL-like queries

 find product pairs sold together in stores in Chicago

this year
 Dimension/level constraint
 in relevance to region, price, brand, customer category

 Rule (or pattern) constraint

 small sales (price < $10) triggers big sales (sum >

$200)
 Interestingness constraint
 strong rules: min_support  3%, min_confidence 

60%
75
 Meta-Rule Guided Mining
 Meta-rule can be in the rule form with partially instantiated predicates
and constants
P1(X, Y) ^ P2(X, W) => buys(X, “iPad”)
 The resulting rule derived can be
age(X, “15-25”) ^ profession(X, “student”) => buys(X, “iPad”)
 In general, it can be in the form of
P1 ^ P2 ^ … ^ Pl => Q1 ^ Q2 ^ … ^ Qr
 Method to find meta-rules
 Find frequent (l+r) predicates (based on min-support threshold)
 Push constants deeply when possible into the mining process (see
the remaining discussions on constraint-push techniques)
 Use confidence, correlation, and other measures when possible
76
 Constraint-Based Frequent Pattern Mining
 Pattern space pruning constraints
 Anti-monotonic: If constraint c is violated, its further mining can
be terminated
 Monotonic: If c is satisfied, no need to check c again
 Succinct: c must be satisfied, so one can start with the data sets
satisfying c
 Convertible: c is not monotonic nor anti-monotonic, but it can be
converted into it if items in the transaction can be properly
ordered
 Data space pruning constraint
 Data succinct: Data space can be pruned at the initial pattern
mining process
 Data anti-monotonic: If a transaction t does not satisfy c, t can be
pruned from its further mining
77
Pattern Space Pruning with Anti-Monotonicity Constraints

TDB (min_sup=2)
 A constraint C is anti-monotone if the super TID Transaction
pattern satisfies C, all of its sub-patterns do so 10 a, b, c, d, f
too
20 b, c, d, f, g, h
 In other words, anti-monotonicity: If an itemset 30 a, c, d, e, f
S violates the constraint, so does any of its 40 c, e, f, g
superset
Item Profit
 Ex. 1. sum(S.price)  v is anti-monotone
a 40
 Ex. 2. range(S.profit)  15 is anti-monotone
b 0
 Itemset ab violates C
c -20
 So does every superset of ab
d 10
 Ex. 3. sum(S.Price)  v is not anti-monotone
e -30
 Ex. 4. support count is anti-monotone: core f 30
property used in Apriori
g 20
h -10 78
 Pattern Space Pruning with Monotonicity Constraints
TDB (min_sup=2)
TID Transaction
 A constraint C is monotone if the pattern
satisfies C, we do not need to check C in 10 a, b, c, d, f
subsequent mining 20 b, c, d, f, g, h
30 a, c, d, e, f
 Alternatively, monotonicity: If an itemset S
40 c, e, f, g
satisfies the constraint, so does any of its
superset Item Profit
 Ex. 1. sum(S.Price)  v is monotone a 40
b 0
 Ex. 2. min(S.Price)  v is monotone
c -20
 Ex. 3. C: range(S.profit)  15
d 10
 Itemset ab satisfies C
e -30
 So does every superset of ab f 30
g 20
h -10 79
 Data Space Pruning with Data Anti-monotonicity
TDB (min_sup=2)
 A constraint c is data anti-monotone if for a pattern TID Transaction
p cannot satisfy a transaction t under c, p’s 10 a, b, c, d, f, h
superset cannot satisfy t under c either 20 b, c, d, f, g, h
 The key for data anti-monotone is recursive data 30 b, c, d, f, g
reduction 40 c, e, f, g
 Ex. 1. sum(S.Price)  v is data anti-monotone Item Profit
 Ex. 2. min(S.Price)  v is data anti-monotone a 40
 Ex. 3. C: range(S.profit)  25 is data anti- b 0

monotone c -20
d -15
 Itemset {b, c}’s projected DB:
e -30
 T10’: {d, f, h}, T20’: {d, f, g, h}, T30’: {d, f, g}
f -10
 since C cannot satisfy T10’, T10’ can be pruned
g 20
h -5 80
 Pattern Space Pruning with Succinctness

 Succinctness:
 Given A1, the set of items satisfying a succinctness
constraint C, then any set S satisfying C is based on
A1 , i.e., S contains a subset belonging to A1
 Idea: 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)  v is succinct
 sum(S.Price)  v is not succinct
 Optimization: If C is succinct, C is pre-counting
pushable 81
 Naïve Algorithm: Apriori + Constraint
Database D itemset sup.
L1 itemset sup.
TID Items C1 {1} 2 {1} 2
100 134 {2} 3 {2} 3
200 235 Scan D {3} 3 {3} 3
300 1235 {4} 1 {5} 3
400 25 {5} 3
C2 itemset sup C2 itemset
L2 itemset sup {1 2} 1 Scan D {1 2}
{1 3} 2 {1 3} 2 {1 3}
{2 3} 2 {1 5} 1 {1 5}
{2 3} 2 {2 3}
{2 5} 3
{2 5} 3 {2 5}
{3 5} 2
{3 5} 2 {3 5}
C3 itemset Scan D L3 itemset sup Constraint:
{2 3 5} {2 3 5} 2 Sum{S.price} < 5
82
Constrained Apriori : Push a Succinct Constraint Deep
Database D itemset sup.
L1 itemset sup.
TID Items C1 {1} 2 {1} 2
100 134 {2} 3 {2} 3
200 235 Scan D {3} 3 {3} 3
300 1235 {4} 1 {5} 3
400 25 {5} 3
C2 itemset sup C2 itemset
L2 itemset sup {1 2}
{1 2} 1 Scan D
{1 3} 2 {1 3} 2 {1 3}
not immediately
{1 5} 1 {1 5} to be used
{2 3} 2
{2 3} 2 {2 3}
{2 5} 3
{2 5} 3 {2 5}
{3 5} 2 {3 5}
{3 5} 2
C3 itemset Scan D L3 itemset sup Constraint:
{2 3 5} {2 3 5} 2 min{S.price } <= 1
83
Constrained FP-Growth: Push a Succinct Constraint
Deep

TID Items TID Items

100 134 100 13
200 235
Remove
200 235
300 1235 infrequent 300 1235 FP-Tree
400 25 length 1
400 25

1-Projected DB
TID Items
100 3 4 No Need to project on 2, 3, or 5
300 2 3 5
Constraint:
min{S.price } <= 1
84
 Constrained FP-Growth: Push a Data Anti-
monotonic Constraint Deep
Remove from data
TID Items TID Items
100 134 100 1 3
200 235 300 1 3
300 1235 FP-Tree
400 25

Single branch, we are done

Constraint:
min{S.price } <= 1
85
 TID Transaction
Constrained FP-Growth: Push a Data
10 a, b, c, d, f, h
Anti-monotonic Constraint Deep 20 b, c, d, f, g, h
30 b, c, d, f, g
TID Transaction
40 a, c, e, f, g
10 a, b, c, d, f, h
20 b, c, d, f, g, h Item Profit

30 b, c, d, f, g FP-Tree a 40
40 a, c, e, f, g b 0
c -20
B-Projected DB Recursive
Data
TID Transaction Pruning
d -15
10 a, c, d, f, h e -30
20 c, d, f, g, h B
f -10
30 c, d, f, g FP-Tree g 20
h -5
Single branch: Constraint:
range{S.price } > 25
bcdfg: 2 min_sup >= 2

86
 Convertible Constraints: Ordering Data in
Transactions
TDB (min_sup=2)
TID Transaction
 Convert tough constraints into anti-
10 a, b, c, d, f
monotone or monotone by properly
20 b, c, d, f, g, h
ordering items 30 a, c, d, e, f
 Examine C: avg(S.profit)  25 40 c, e, f, g
 Order items in value-descending Item Profit
order a 40
 <a, f, g, d, b, h, c, e> b 0
c -20
 If an itemset afb violates C d 10
 So does afbh, afb* e -30
f 30
 It becomes anti-monotone!
g 20
h -10
87
 Strongly Convertible Constraints

 avg(X)  25 is convertible anti-monotone w.r.t.

item value descending order R: <a, f, g, d, b,
h, c, e> Item Profit
 If an itemset af violates a constraint C, so
a 40
does every itemset with af as prefix, such as b 0
afd
c -20
 avg(X)  25 is convertible monotone w.r.t. item d 10
value ascending order R-1: <e, c, h, b, d, g, f,
e -30
a>
 If an itemset d satisfies a constraint C, so
f 30
g 20
does itemsets df and dfa, which having d as
a prefix h -10
 Thus, avg(X)  25 is strongly convertible
88
 Can Apriori Handle Convertible Constraints?

 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
Item Value
pruning based on the constraint can be made
a 40
 Itemset df violates constraint C: avg(X) >= b 0
25 c -20
 Since adf satisfies C, Apriori needs df to d 10

assemble adf, df cannot be pruned e -30

f 30
 But it can be pushed into frequent-pattern
g 20
growth framework! h -10

89
 Pattern Space Pruning w. Convertible Constraints
Item Value
 C: avg(X) >= 25, min_sup=2
a 40
 List items in every transaction in value
f 30
descending order R: <a, f, g, d, b, h, c, e>
g 20
 C is convertible anti-monotone w.r.t. R
d 10
 Scan TDB once b 0
 remove infrequent items
h -10
 Item h is dropped
c -20
 Itemsets a and f are good, …
e -30
 Projection-based mining TDB (min_sup=2)
 Imposing an appropriate order on item TID Transaction

projection 10 a, f, d, b, c
 Many tough constraints can be converted into 20 f, g, d, b, c
(anti)-monotone 30 a, f, d, c, e
40 f, g, h, c, e
90
 Handling Multiple Constraints

 Different constraints may require different or even

conflicting item-ordering
 If there exists an order R s.t. both C1 and C2 are
convertible w.r.t. R, then there is no conflict between
the two convertible constraints
 If 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
91
 What Constraints Are Convertible?

Convertible anti- Convertible Strongly

Constraint monotone monotone convertible

avg(S)  ,  v Yes Yes Yes

median(S)  ,  v Yes Yes Yes
sum(S)  v (items could be of any value,
Yes No No
v  0)
sum(S)  v (items could be of any value,
No Yes No
v  0)
sum(S)  v (items could be of any value,
No Yes No
v  0)
sum(S)  v (items could be of any value,
Yes No No
v  0)
……

92
Constraint-Based Mining — A General Picture

Constraint Anti-monotone Monotone Succinct

vS no yes yes
SV no yes yes

SV yes no yes

min(S)  v no yes yes

min(S)  v yes no yes

max(S)  v yes no yes

max(S)  v no yes yes

count(S)  v yes no weakly

count(S)  v no yes weakly

sum(S)  v ( a  S, a  0 ) yes no no
sum(S)  v ( a  S, a  0 ) no yes no

range(S)  v yes no no
range(S)  v no yes no

avg(S)  v,   { , ,  } convertible convertible no

support(S)   yes no no

support(S)   no yes no

93
 Chapter 7 : Advanced Frequent Pattern Mining

 Pattern Mining: A Road Map

 Pattern Mining in Multi-Level, Multi-Dimensional Space

 Constraint-Based Frequent Pattern Mining

 Mining High-Dimensional Data and Colossal Patterns

 Mining Compressed or Approximate Patterns

 Pattern Exploration and Application

 Summary

94
 Mining Colossal Frequent Patterns
 F. Zhu, X. Yan, J. Han, P. S. Yu, and H. Cheng, “Mining Colossal
Frequent Patterns by Core Pattern Fusion”, ICDE'07.
 We have many algorithms, but can we mine large (i.e., colossal)
patterns? ― such as just size around 50 to 100? Unfortunately, not!
 Why not? ― the curse of “downward closure” of frequent patterns
 The “downward closure” property
 Any sub-pattern of a frequent pattern is frequent.
 Example. If (a1, a2, …, a100) is frequent, then a1, a2, …, a100, (a1,
a2), (a1, a3), …, (a1, a100), (a1, a2, a3), … are all frequent! There
are about 2100 such frequent itemsets!
 No matter using breadth-first search (e.g., Apriori) or depth-first
search (FPgrowth), we have to examine so many patterns
 Thus the downward closure property leads to explosion!
95
 Colossal Patterns: A Motivating Example
Let’s make a set of 40 transactions Closed/maximal patterns may
T1 = 1 2 3 4 ….. 39 40 partially alleviate the problem but not
T2 = 1 2 3 4 ….. 39 40 really solve it: We often need to
: . mine scattered large patterns!
: . Let the minimum support threshold
: . σ= 20
: .  40 
There are   frequent patterns of
T40=1 2 3 4 ….. 39 size 20  20 
Then 40
delete the items on the diagonal
Each is closed and maximal
T1 = 2 3 4 ….. 39 40
T2 = 1 3 4 ….. 39 40 # patterns =  n  2n
: .    2 / 
: .  n / 2 n
: .
The size of the answer set is
: . exponential to n
T40=1 2 3 4 …… 39
96
 Colossal Pattern Set: Small but Interesting

 It is often the case that

only a small number of
patterns are colossal,
i.e., of large size

 Colossal patterns are

usually attached with
greater importance than
those of small pattern
sizes

97
 Mining Colossal Patterns: Motivation and
Philosophy
 Motivation: Many real-world tasks need mining colossal patterns
 Micro-array analysis in bioinformatics (when support is low)

 Biological sequence patterns

 Biological/sociological/information graph pattern mining

 No hope for completeness

 If the mining of mid-sized patterns is explosive in size, there is no
hope to find colossal patterns efficiently by insisting “complete set”
mining philosophy
 Jumping out of the swamp of the mid-sized results
 What we may develop is a philosophy that may jump out of the
swamp of mid-sized results that are explosive in size and jump to
reach colossal patterns
 Striving for mining almost complete colossal patterns
 The key is to develop a mechanism that may quickly reach colossal
patterns and discover most of them
98
 Alas, A Show of Colossal Pattern Mining!

Let the min-support threshold σ= 20

T1 = 2 3 4 ….. 39 40  40 
T2 = 1 3 4 ….. 39 40 Then there are   closed/maximal
 20 
 
: . frequent patterns of size 20
: .
: . However, there is only one with size
: . greater than 20, (i.e., colossal):
T40=1 2 3 4 …… 39
α= {41,42,…,79} of size 39
T41= 41 42 43 ….. 79
T42= 41 42 43 ….. 79
: . The
Theexisting
existingfastest
fastest mining
miningalgorithms
algorithms
: . (e.g.,
(e.g., FPClose,
FPClose, LCM)
LCM) fail
failto
tocomplete
complete
T60= 41 42 43 … 79
running
running
Our
Our algorithm
algorithm outputs
outputsthis
thiscolossal
colossal
pattern
patternininseconds
seconds
99
 Methodology of Pattern-Fusion Strategy
 Pattern-Fusion traverses the tree in a bounded-breadth way
 Always pushes down a frontier of a bounded-size candidate
pool
 Only a fixed number of patterns in the current candidate pool
will be used as the starting nodes to go down in the pattern tree
― thus avoids the exponential search space
 Pattern-Fusion identifies “shortcuts” whenever possible
 Pattern growth is not performed by single-item addition but by
leaps and bounded: agglomeration of multiple patterns in the
pool
 These shortcuts will direct the search down the tree much more
rapidly towards the colossal patterns
100
 Observation: Colossal Patterns and Core Patterns

Transaction Database D
A colossal pattern α
α D

α1 Dαk
α2
D
Dα1
α
Dα2

αk

Subpatterns α1 to αk cluster tightly around the colossal pattern α by

sharing a similar support. We call such subpatterns core patterns of α

101
 Robustness of Colossal Patterns
 Core Patterns
Intuitively, for a frequent pattern α, a subpattern β is a τ-core
pattern of α if β shares a similar support set with α, i.e.,
| D |
 0  1
| D |

where τ is called the core ratio

 Robustness of Colossal Patterns
A colossal pattern is robust in the sense that it tends to have much
more core patterns than small patterns

102
 Example: Core Patterns
 A colossal pattern has far more core patterns than a small-sized pattern
 A colossal pattern has far more core descendants of a smaller size c
 A random draw from a complete set of pattern of size c would more
likely to pick a core descendant of a colossal pattern
 A colossal pattern can be generated by merging a set of core patterns

Transaction (# of Core Patterns (τ = 0.5)

Ts)
(abe) (100) (abe), (ab), (be), (ae), (e)
(bcf) (100) (bcf), (bc), (bf)
(acf) (100) (acf), (ac), (af)

(abcef) (100) (ab), (ac), (af), (ae), (bc), (bf), (be) (ce), (fe), (e),
(abc), (abf), (abe), (ace), (acf), (afe), (bcf), (bce),
(bfe), (cfe), (abcf), (abce), (bcfe), (acfe), (abfe),
(abcef)
103
 Colossal Patterns Correspond to Dense Balls

 Due to their robustness,

colossal patterns correspond to
dense balls
 Ω( 2^d) in population
 A random draw in the pattern
space will hit somewhere in the
ball with high probability

105
 Idea of Pattern-Fusion Algorithm

 Generate a complete set of frequent patterns up to a

small size
 Randomly pick a pattern β, and β has a high probability
to be a core-descendant of some colossal pattern α
 Identify all α’s descendants in this complete set, and
merge all of them ― This would generate a much larger
core-descendant of α
 In the same fashion, we select K patterns. This set of
larger core-descendants will be the candidate pool for the
next iteration

106
 Pattern-Fusion: The Algorithm
 Initialization (Initial pool): Use an existing algorithm to
mine all frequent patterns up to a small size, e.g., 3
 Iteration (Iterative Pattern Fusion):
 At each iteration, k seed patterns are randomly picked

from the current pattern pool

 For each seed pattern thus picked, we find all the

patterns within a bounding ball centered at the seed

pattern
 All these patterns found are fused together to generate

a set of super-patterns. All the super-patterns thus

generated form a new pool for the next iteration
 Termination: when the current pool contains no more
than K patterns at the beginning of an iteration
107
 Why Is Pattern-Fusion Efficient?
 A bounded-breadth pattern
tree traversal
 It avoids explosion in

mining mid-sized ones

 Randomness comes to help

to stay on the right path

 Ability to identify “short-cuts”
and take “leaps”
 fuse small patterns

together in one step to

generate new patterns of
significant sizes
 Efficiency

108
Pattern-Fusion Leads to Good Approximation

 Gearing toward colossal patterns

 The larger the pattern, the greater the chance it will
be generated
 Catching outliers
 The more distinct the pattern, the greater the chance
it will be generated

109
 Experimental Setting

 Synthetic data set

 Diagn an n x (n-1) table where ith row has integers from 1 to n
except i. Each row is taken as an itemset. min_support is n/2.
 Real data set
 Replace: A program trace data set collected from the “replace”
program, widely used in software engineering research
 ALL: A popular gene expression data set, a clinical data on ALL-AML
leukemia (www.broad.mit.edu/tools/data.html).
 Each item is a column, representing the activitiy level of
gene/protein in the same
 Frequent pattern would reveal important correlation between
gene expression patterns and disease outcomes

110
 Experiment Results on Diagn
 LCM run time increases
exponentially with pattern
size n
 Pattern-Fusion finishes
efficiently
 The approximation error of
Pattern-Fusion (with min-sup
20) in comparison with the
complete set) is rather close
to uniform sampling (which
randomly picks K patterns
from the complete answer
set)

111
 Experimental Results on ALL
 ALL: A popular gene expression data set with 38
transactions, each with 866 columns
 There are 1736 items in total

 The table shows a high frequency threshold of 30

112
 Experimental Results on REPLACE

 REPLACE
 A program trace data set, recording 4395 calls

and transitions
 The data set contains 4395 transactions with

57 items in total
 With support threshold of 0.03, the largest

patterns are of size 44

 They are all discovered by Pattern-Fusion with

different settings of K and τ, when started with

an initial pool of 20948 patterns of size <=3

113
 Experimental Results on REPLACE
 Approximation error when
compared with the complete
mining result
 Example. Out of the total 98
patterns of size >=42, when
K=100, Pattern-Fusion returns
80 of them
 A good approximation to the
colossal patterns in the sense
that any pattern in the
complete set is on average at
most 0.17 items away from one
of these 80 patterns

114
 Chapter 7 : Advanced Frequent Pattern Mining

 Pattern Mining: A Road Map

 Pattern Mining in Multi-Level, Multi-Dimensional Space

 Constraint-Based Frequent Pattern Mining

 Mining High-Dimensional Data and Colossal Patterns

 Mining Compressed or Approximate Patterns

 Pattern Exploration and Application

 Summary

115
 Mining Compressed Patterns: δ-clustering
 Why compressed patterns? ID Item-Sets Support

P1 {38,16,18,12} 205227
 too many, but less meaningful
P2 {38,16,18,12,17} 205211
 Pattern distance measure P3 {39,38,16,18,12,17 101758
}
P4 {39,16,18,12,17} 161563

P5 {39,16,18,12} 161576
 δ-clustering: For each pattern P,  Closed frequent pattern
find all patterns which can be  Report P1, P2, P3, P4, P5
expressed by P and their distance  Emphasize too much on
to P are within δ (δ-cover) support
 All patterns in the cluster can be  no compression

represented by P  Max-pattern, P3: info loss

 Xin et al., “Mining Compressed  A desirable output: P2, P3,
Frequent-Pattern Sets”, VLDB’05 P4

116
 Redundancy-Award Top-k Patterns
 Why redundancy-aware top-k patterns?
 Desired patterns: high
significance & low
redundancy
 Propose the MMS
(Maximal Marginal
Significance) for
measuring the
combined significance
of a pattern set
 Xin et al., Extracting
Redundancy-Aware
Top-K Patterns, KDD’06
117
 Chapter 7 : Advanced Frequent Pattern Mining

 Pattern Mining: A Road Map

 Pattern Mining in Multi-Level, Multi-Dimensional Space

 Constraint-Based Frequent Pattern Mining

 Mining High-Dimensional Data and Colossal Patterns

 Mining Compressed or Approximate Patterns

 Pattern Exploration and Application

 Summary
118
How to Understand and Interpret Patterns?

 Do they all make sense?

diaper beer  What do they mean?
 How are they useful?
female sterile (2) tekele

morphological info. and simple statistics

Semantic Information

Not all frequent patterns are useful, only meaningful ones …

Annotate patterns with semantic information

 A Dictionary Analogy
Word: “pattern” – from Merriam-Webster

Non-semantic info.

Definitions indicating
semantics

Examples of Usage
Synonyms

Related Words
Semantic Analysis with Context Models

 Task1: Model the context of a frequent pattern

Based on the Context Model…
 Task2: Extract strongest context indicators
 Task3: Extract representative transactions
 Task4: Extract semantically similar patterns
 Annotating DBLP Co-authorship & Title Pattern
Database: Frequent Patterns
Authors Title

Substructure Similarity Search P1: { x_yan, j_han }

X.Yan, P. Yu, J. Han
in Graph Databases
Frequent Itemset
… …

… … P2: “substructure search”

Semantic Annotations
Pattern { x_yan, j_han} Context Units
Non Sup = …
< { p_yu, j_han}, { d_xin }, … , “graph pattern”,
CI {p_yu}, graph pattern, … … “substructure similarity”, … >
Trans. gSpan: graph-base……
SSPs { j_wang }, {j_han, p_yu}, …

Pattern = {xifeng_yan, jiawei_han} Annotation Results:

Context Indicator (CI) graph; {philip_yu}; mine close; graph pattern; sequential pattern; …
Representative > gSpan: graph-base substructure pattern mining;
Transactions (Trans) > mining close relational graph connect constraint; …
Semantically Similar {jiawei_han, philip_yu}; {jian_pei, jiawei_han}; {jiong_yang, philip_yu,
Patterns (SSP) wei_wang}; …
 Chapter 7 : Advanced Frequent Pattern Mining

 Pattern Mining: A Road Map

 Pattern Mining in Multi-Level, Multi-Dimensional Space

 Constraint-Based Frequent Pattern Mining

 Mining High-Dimensional Data and Colossal Patterns

 Mining Compressed or Approximate Patterns

 Pattern Exploration and Application

 Summary
123
 Summary
 Roadmap: Many aspects & extensions on pattern mining
 Mining patterns in multi-level, multi dimensional space
 Mining rare and negative patterns
 Constraint-based pattern mining
 Specialized methods for mining high-dimensional data
and colossal patterns
 Mining compressed or approximate patterns
 Pattern exploration and understanding: Semantic
annotation of frequent patterns
124
 Ref: Mining Multi-Level and Quantitative Rules
 Y. Aumann and Y. Lindell. A Statistical Theory for Quantitative Association
Rules, KDD'99
 T. Fukuda, Y. Morimoto, S. Morishita, and T. Tokuyama. Data mining using
two-dimensional optimized association rules: Scheme, algorithms, and
visualization. SIGMOD'96.
 J. Han and Y. Fu. Discovery of multiple-level association rules from large
databases. VLDB'95.
 R.J. Miller and Y. Yang. Association rules over interval data. SIGMOD'97.
 R. Srikant and R. Agrawal. Mining generalized association rules. VLDB'95.
 R. Srikant and R. Agrawal. Mining quantitative association rules in large
relational tables. SIGMOD'96.
 K. Wang, Y. He, and J. Han. Mining frequent itemsets using support
constraints. VLDB'00
 K. Yoda, T. Fukuda, Y. Morimoto, S. Morishita, and T. Tokuyama. Computing
optimized rectilinear regions for association rules. KDD'97.
125
 Ref: Mining Other Kinds of Rules
 F. Korn, A. Labrinidis, Y. Kotidis, and C. Faloutsos. Ratio rules: A new
paradigm for fast, quantifiable data mining. VLDB'98
 Y. Huhtala, J. Kärkkäinen, P. Porkka, H. Toivonen. Efficient Discovery of
Functional and Approximate Dependencies Using Partitions. ICDE’98.
 H. V. Jagadish, J. Madar, and R. Ng. Semantic Compression and Pattern
Extraction with Fascicles. VLDB'99
 B. Lent, A. Swami, and J. Widom. Clustering association rules. ICDE'97.
 R. Meo, G. Psaila, and S. Ceri. A new SQL-like operator for mining
association rules. VLDB'96.
 A. Savasere, E. Omiecinski, and S. Navathe. Mining for strong negative
associations in a large database of customer transactions. ICDE'98.
 D. Tsur, J. D. Ullman, S. Abitboul, C. Clifton, R. Motwani, and S. Nestorov.
Query flocks: A generalization of association-rule mining. SIGMOD'98.

126
 Ref: Constraint-Based Pattern Mining
 R. Srikant, Q. Vu, and R. Agrawal. Mining association rules with item
constraints. KDD'97
 R. Ng, L.V.S. Lakshmanan, J. Han & A. Pang. Exploratory mining and pruning
optimizations of constrained association rules. SIGMOD’98
 G. Grahne, L. Lakshmanan, and X. Wang. Efficient mining of constrained
correlated sets. ICDE'00
 J. Pei, J. Han, and L. V. S. Lakshmanan. Mining Frequent Itemsets with
Convertible Constraints. ICDE'01
 J. Pei, J. Han, and W. Wang, Mining Sequential Patterns with Constraints in
Large Databases, CIKM'02
 F. Bonchi, F. Giannotti, A. Mazzanti, and D. Pedreschi. ExAnte: Anticipated
Data Reduction in Constrained Pattern Mining, PKDD'03
 F. Zhu, X. Yan, J. Han, and P. S. Yu, “gPrune: A Constraint Pushing Framework
for Graph Pattern Mining”, PAKDD'07
127
 Ref: Mining Sequential Patterns
 X. Ji, J. Bailey, and G. Dong. Mining minimal distinguishing subsequence patterns with
gap constraints. ICDM'05
 H. Mannila, H Toivonen, and A. I. Verkamo. Discovery of frequent episodes in event
sequences. DAMI:97.
 J. Pei, J. Han, H. Pinto, Q. Chen, U. Dayal, and M.-C. Hsu. PrefixSpan: Mining Sequential
Patterns Efficiently by Prefix-Projected Pattern Growth. ICDE'01.
 R. Srikant and R. Agrawal. Mining sequential patterns: Generalizations and
performance improvements. EDBT’96.
 X. Yan, J. Han, and R. Afshar. CloSpan: Mining Closed Sequential Patterns in Large
Datasets. SDM'03.
 M. Zaki. SPADE: An Efficient Algorithm for Mining Frequent Sequences. Machine
Learning:01.

128
 Mining Graph and Structured Patterns
 A. Inokuchi, T. Washio, and H. Motoda. An apriori-based algorithm for
mining frequent substructures from graph data. PKDD'00
 M. Kuramochi and G. Karypis. Frequent Subgraph Discovery. ICDM'01.
 X. Yan and J. Han. gSpan: Graph-based substructure pattern mining.
ICDM'02
 X. Yan and J. Han. CloseGraph: Mining Closed Frequent Graph Patterns.
KDD'03
 X. Yan, P. S. Yu, and J. Han. Graph indexing based on discriminative frequent
structure analysis. ACM TODS, 30:960–993, 2005
 X. Yan, F. Zhu, P. S. Yu, and J. Han. Feature-based substructure similarity
search. ACM Trans. Database Systems, 31:1418–1453, 2006

129
 Ref: Mining Spatial, Spatiotemporal, Multimedia Data

 H. Cao, N. Mamoulis, and D. W. Cheung. Mining frequent spatiotemporal

sequential patterns. ICDM'05
 D. Gunopulos and I. Tsoukatos. Efficient Mining of Spatiotemporal Patterns.
SSTD'01
 K. Koperski and J. Han, Discovery of Spatial Association Rules in Geographic
Information Databases, SSD’95
 H. Xiong, S. Shekhar, Y. Huang, V. Kumar, X. Ma, and J. S. Yoo. A framework
for discovering co-location patterns in data sets with extended spatial
objects. SDM'04
 J. Yuan, Y. Wu, and M. Yang. Discovery of collocation patterns: From visual
words to visual phrases. CVPR'07
 O. R. Zaiane, J. Han, and H. Zhu, Mining Recurrent Items in Multimedia with
Progressive Resolution Refinement. ICDE'00

130
 Ref: Mining Frequent Patterns in Time-Series Data

 B. Ozden, S. Ramaswamy, and A. Silberschatz. Cyclic association rules. ICDE'98.

 J. Han, G. Dong and Y. Yin, Efficient Mining of Partial Periodic Patterns in Time Series
Database, ICDE'99.
 J. Shieh and E. Keogh. iSAX: Indexing and mining terabyte sized time series. KDD'08
 B.-K. Yi, N. Sidiropoulos, T. Johnson, H. V. Jagadish, C. Faloutsos, and A. Biliris. Online
Data Mining for Co-Evolving Time Sequences. ICDE'00.
 W. Wang, J. Yang, R. Muntz. TAR: Temporal Association Rules on Evolving Numerical
Attributes. ICDE’01.
 J. Yang, W. Wang, P. S. Yu. Mining Asynchronous Periodic Patterns in Time Series Data.
TKDE’03
 L. Ye and E. Keogh. Time series shapelets: A new primitive for data mining. KDD'09

131
 Ref: FP for Classification and Clustering
 G. Dong and J. Li. Efficient mining of emerging patterns: Discovering
trends and differences. KDD'99.
 B. Liu, W. Hsu, Y. Ma. Integrating Classification and Association Rule
Mining. KDD’98.
 W. Li, J. Han, and J. Pei. CMAR: Accurate and Efficient Classification Based
on Multiple Class-Association Rules. ICDM'01.
 H. Wang, W. Wang, J. Yang, and P.S. Yu. Clustering by pattern similarity in
large data sets. SIGMOD’ 02.
 J. Yang and W. Wang. CLUSEQ: efficient and effective sequence clustering.
ICDE’03.
 X. Yin and J. Han. CPAR: Classification based on Predictive Association
Rules. SDM'03.
 H. Cheng, X. Yan, J. Han, and C.-W. Hsu, Discriminative Frequent Pattern
Analysis for Effective Classification”, ICDE'07
132
 Ref: Privacy-Preserving FP Mining

 A. Evfimievski, R. Srikant, R. Agrawal, J. Gehrke. Privacy Preserving Mining

of Association Rules. KDD’02.
 A. Evfimievski, J. Gehrke, and R. Srikant. Limiting Privacy Breaches in
Privacy Preserving Data Mining. PODS’03
 J. Vaidya and C. Clifton. Privacy Preserving Association Rule Mining in
Vertically Partitioned Data. KDD’02

133
 Mining Compressed Patterns
 D. Xin, H. Cheng, X. Yan, and J. Han. Extracting redundancy-
aware top-k patterns. KDD'06
 D. Xin, J. Han, X. Yan, and H. Cheng. Mining compressed
frequent-pattern sets. VLDB'05
 X. Yan, H. Cheng, J. Han, and D. Xin. Summarizing itemset
patterns: A profile-based approach. KDD'05

134
 Mining Colossal Patterns
 F. Zhu, X. Yan, J. Han, P. S. Yu, and H. Cheng. Mining colossal
frequent patterns by core pattern fusion. ICDE'07
 F. Zhu, Q. Qu, D. Lo, X. Yan, J. Han. P. S. Yu, Mining Top-K Large
Structural Patterns in a Massive Network. VLDB’11

135
 Ref: FP Mining from Data Streams
 Y. Chen, G. Dong, J. Han, B. W. Wah, and J. Wang. Multi-Dimensional
Regression Analysis of Time-Series Data Streams. VLDB'02.
 R. M. Karp, C. H. Papadimitriou, and S. Shenker. A simple algorithm for
finding frequent elements in streams and bags. TODS 2003.
 G. Manku and R. Motwani. Approximate Frequency Counts over Data
Streams. VLDB’02.
 A. Metwally, D. Agrawal, and A. El Abbadi. Efficient computation of frequent
and top-k elements in data streams. ICDT'05

136
 Ref: Freq. Pattern Mining Applications

 T. Dasu, T. Johnson, S. Muthukrishnan, and V. Shkapenyuk. Mining Database Structure; or

How to Build a Data Quality Browser. SIGMOD'02
 M. Khan, H. Le, H. Ahmadi, T. Abdelzaher, and J. Han. DustMiner: Troubleshooting
interactive complexity bugs in sensor networks., SenSys'08
 Z. Li, S. Lu, S. Myagmar, and Y. Zhou. CP-Miner: A tool for finding copy-paste and related
bugs in operating system code. In Proc. 2004 Symp. Operating Systems Design and
Implementation (OSDI'04)
 Z. Li and Y. Zhou. PR-Miner: Automatically extracting implicit programming rules and
detecting violations in large software code. FSE'05
 D. Lo, H. Cheng, J. Han, S. Khoo, and C. Sun. Classification of software behaviors for failure
detection: A discriminative pattern mining approach. KDD'09
 Q. Mei, D. Xin, H. Cheng, J. Han, and C. Zhai. Semantic annotation of frequent patterns.
ACM TKDD, 2007.
 K. Wang, S. Zhou, J. Han. Profit Mining: From Patterns to Actions. EDBT’02.

137