Data Mining: Concepts and Techniques (3 ed.) — Chapter 6 — Jiawei Han, Micheline Kamber, and Jian Pei University of Illinois at Urbana-Champaign & Simon Fraser University ©2011 Han, Kamber & Pei. All rights reserved.Chapter 5: Mining Frequent Patterns, Association and Correlations: Basic Concepts and Methods @ Basic Concepts @ m@ Frequent Itemset Mining Methods m@ Which Patterns Are Interesting?—Pattern Evaluation Methods m@ Summary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.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Basic Concepts: Frequent Patterns 10 Beer, Nuts, Diaper 20 Beer, Coffee, Diaper 30 Beer, Diaper, Eggs 40 Nuts, Eggs, Milk 50 Nuts, Coffee, Diaper, Eggs, Milk Customer Customer buys diaper Customer buys beer itemset: A set of one or more items k-itemset X = {Xq, ..., Xk} (absolute) support, or, support count of X: Frequency or occurrence of an itemset X (relative) support, s, is the 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 thresholdBasic Concepts: Association Rules 10 Beer, Nuts, Diaper 20 Beer, Coffee, Diaper 30 Beer, Diaper, Eggs 40 Nuts, Eggs, Milk 50 | Nuts, Coffee, Diaper, Eggs, Mik Customer Customer ‘Customer buys beer = Find all the rules X > Ywith minimum support and confidence = support, s, probability that a transaction contains X U Y = confidence, c conditional probability that a transaction having X also contains Y Let minsup = 50%, minconf = 50% Freq. Pat.: Beer:3, Nuts:3, Diaper:4, Eggs:3, {Beer, Diaper}:3 = Association rules: (many more!) » Beer > Diaper (60%, 100%) = Diaper > Beer (60%, 75%)Closed Patterns and Max-Patterns A long pattern contains a combinatorial number of sub- patterns, e.g., {@1, ..-, Aygo} Contains (499) + (4002) +. + (125%0°) = 210°- 1 = 1.27*10%° 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 > X, 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 > X (proposed by Bayardo @ SIGMOD’98) Closed pattern is a lossless compression of freq. patterns = Reducing the # of patterns and rulesClosed Patterns and Max-Patterns Exercise. DB = {) < Ay, «+, Aso>} = Min_sup = 1. What is the set of closed itemset? = : 1 m : 2 What is the set of max-pattern? m : 1 What is the set of all patterns?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: M" where M: # distinct items, and N: max length of transactions = The worst case complexty vs. the expected probability = Ex. Suppose Walmart has 10‘ kinds of products » The chance to pick up one product 10+ « The chance to pick up a particular set of 10 products: ~10-4° « What is the chance this particular set of 10 products to be frequent 10° times in 10° transactions?Chapter 5: Mining Frequent Patterns, Association and Correlations: Basic Concepts and Methods @ Basic Concepts m@ Frequent Itemset Mining Methods @ m@ Which Patterns Are Interesting?—Pattern Evaluation Methods m@ Summary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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)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The Apriori Algorithm—An Example Si = 2 [Htemset [sup | Database TPB >™* {Ay 2] 7 Cc, {BY 3 “1 10 | ACD {} 3 20 BGE 18 scan 30 | ABGE {E} 3 40 BE Cc; Ly ¢The Apriori Algorithm (Pseudo-Code) C, Candidate itemset of size k L,: frequent itemset of size k L, = {frequent items}; for (k= 1; 4, !=@; k++) do begin C1 = candidates generated from L, for each transaction tin database do increment the count of all candidates in C,,, that are contained in ¢ Ly, = candidates in C,,, with min_support end return U; LyImplementation of Apriori = How to generate candidates? = Step 1: self-joining Ly = Step 2: pruning = Example of Candidate-generation « L3={abc, abd, acd, ace, bcd} = Self-joining: £;*L3 » abcd from abcand abd » acdefrom acdand ace = Pruning: » acde is removed because ade is not in £3 = C= {abcd}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 functior. finds all the candidates contained in a transactionCounting Supports of Candidates Using Hash Tree Subset function 1 aye Transaction: 12356 2,5,8 142356 >. 13+56-~._ 356 367 - 368 124+356--"" tne 134 457 125Candidate Generation: An SQL Implementation = SQL Implementation of candidate generation = Suppose the items in /,.; are listed in an order = Step 1: self-joining Ly, insert into C, select p.itemy p.itemz, ..., p.itemyy q.itemy, from Ly.2 Py har] where p.item,=q.itemy ... qg.itemy., = Step 2: pruning forall itemsets c in C,,do forall (k-1)-subsets s of cdo if (5 is not in L,.,) then delete cfrom C, = 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] p.item,.2=q.itemy.z p.itemy.1<Scalable Frequent Itemset Mining Methods = Apriori: A Candidate Generation-and-Test Approach = Improving the Efficiency of Apriori S = FPGrowth: A Frequent Pattern-Growth Approach = ECLAT: Frequent Pattern Mining with Vertical Data Format a Mining Close Frequent Patterns and Maxpatterns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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 A. Savasere, E. Omiecinski and S. Navathe, VLDB95 DB, + DB, + + DBR = OoB sup,(i) < oDB; —sup,(i) < oDB, sup,(i) < oDB, —_sup(i) < oDBDHP: Reduce the Number of Candidates = A &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} noe L102 | yz, a; 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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 orders 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DIC: Reduce Number of Scans = 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 On 1-itemsets Apriori 2-itemsets Itemset lattice 1-itemsets S. Brin R. Motwani, J. Ullman, and S. Tsur. Dynamic itemset DIC counting and implication rules for market basket data. SIGMOD97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 a Mining Close Frequent Patterns and Maxpatterns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Construct FP-tree from a Transaction Database TID ____Items bought (ordered) frequent items 100 tf 4, cd, g, i, m, p} tf G a, m, p} 200 {a, b,c, fl, m, of tf & a, b, m} . 300 {6h j, 0, w} ; min_support = 3 400 {b, c, k, s, p} {c, b, p} 500 {a, f, ce, 1 p,m, n} {f Ga, m, p} G Header Table 1. Scan DB once, find , frequent 1-itemset (single Item frequency head | ~~ f4 | ale: item pattern) f 4 --47 c 2. Sort frequent items in a 3 frequency descending b 3 order, f-list m 3 Dp 3 3. Scan DB again, construct FP-tree F-list = fea-b-m-p $p:2} m1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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 pto form p& conditional pattern base Header Table Item frequency head | ---tf4|->\e:1 Conditional pattern bases f 4 = 4" item cond. pattern base e 4 5 - 5 « (3 b 3 a fe3 m 3 b fea:l, fel, c:1 P 3 m ——_fea:2, feab:1 Pp Sfeam:2, cb:1 30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 pattern base Header Table Item frequency head. he frequent items of the m-conditional pattern base: fea:2, feab:1 All frequent patterns relate to m y a> Imm, am, I ‘fem, fam, cam, c:3feam I a:3 m-conditional FP-tree >Recursion: Mining Each Conditional FP-tree u | 7 Cond. pattern base of “am”: (fc:3) £3 13 e:3 1 am-conditional FP-tree o3 G ! Cond. pattern base of “cm”: (f:3) | a:3 f3 m-conditional FP -tree em-conditional FP-tree 3 Cond. pattern base of “cam”: (f:3) 3 cam-conditional FP-tree 32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 4" = Concatenation of the mining results of the two aainy parts a3:n3 /\ » ¢ F Crk, ayn, / \ bymy, 1K Y= 1 + bum Cyrky ay:ny Giko Gtks ' a3:n; Cotky Corks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)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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Partition-Based Projection = Parallel projection needs a lot of disk space fcamp = Partition projection saves it foPerformance of FPGrowth in Large Datasets ‘e | ro \ = D2 FP guts i \ Ot Apiierine 120 s + -02 TreePrejection _n \ 5m \ a= \ = 0 EF \ Data set T25120D10K Data set T25120D 100K :e Eo x 20 za a oo i as 2 2 3 0 05 4 15 2 Support tresolat) Suppor testa (9 FP-Growth vs. Apriori FP-Growth vs. Tree-Projection 38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)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) 40Extension of Pattern Growth Mining Methodology a 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) 41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 & a Mining Close Frequent Patterns and Maxpatterns 42ECLAT: Mining by Exploring Vertical Data Format = Vertical format: t(AB) = {T41, Tos, ...} = 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) c t(Y): transaction having X always has Y = Using diffset to accelerate mining = Only keep track of differences of tids = tX%) = {Ty Ta Ta}, (XY) = {T1, Ts} = Diffset (XY, X) = {12} = Eclat (Zaki et al. @KDD’97) = Mining Closed patterns using vertical format: CHARM (Zaki & Hsiao@SDM’'02) 43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 a Mining Close Frequent Patterns and Maxpatterns &Mining Frequent Closed Patterns: CLOSET Flist: list of all frequent items in support ascending order = Flist: d-a-f-e-c Min_sup=2 Divide search space ~ Items = Patterns having d 2 = Patterns having d but no a, etc. 40 50 Find frequent closed pattern recursively = Every transaction having d also has cfa> cfadisa 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 > 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 = 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 1st scan: find frequent items Tid _| Items 10 |A,B,C,D,E » A,B, C,D,E 20 1B,C,D,E, 24 scan: find support for 30_|A,C,D,F = AB, AC, AD, AE, ABCDE = BC, BD, BE, BCDE 9 __ Potential = CD, CE, CDE,.DE————~" max-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, SIGMOD98CHARM: Mining by Exploring Vertical Data Format Vertical format: t(AB) = {T41, Tas, ...} = 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) c t(Y): transaction having X always has Y Using diffset to accelerate mining = Only keep track of differences of tids = U(X) = {Ti Tar Ts}, (XY) = {T1, Ts} « Diffset (XY, X) = {Tp} 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 =| sal soa stpVisualization of Association Rules $GI/MineSet 3.0Chapter 5: Mining Frequent Patterns, Association and Correlations: Basic Concepts and Methods @ Basic Concepts m@ Frequent Itemset Mining Methods m@ Which Patterns Are Interesting?—Pattern @e Evaluation Methods m@ Summary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 lift = P(AUB) Basketball | Not basketball | Sum (row) ift ~ P(A)P(B) Cereal 2000 1750 3750 Not cereal | 1000 250 1250 2000/5000 lift(B,C)= 3000/5000*3750/5000 ~ 0.89 Tsum(col) | 3000 2000 5000 lift(B, aC) = 1000/5000 =133 3000 /5000*1250/5000Are /ift and y?_ Good Measures of Correlation? Smbol “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 7 proposed (see Tan, Kumar, Sritastava a @KDD'02) 7 Which are good ones? : Y ules Yutes¥ Platetsky-Shpiro'e Cert factor Measure Laplace Cost ‘coherence( accatd) al confdence ds ati lite Collective strength = gee gee Leib ‘nax(PLB|A), P(AIB)) ace ARRProperties of interestingness measures. Note that none of the measures satisfies all the properties. -Invariant Measures ‘Table ipsa ees ' Ta STS Or Eon Or corpo) S| Gecaherkrosa's 1 No | ve | So [ae | vee | No 2 | cca cnte « xe ye Se |e | ie @ | vue ot xe ye | Se fas | Se $ )weeg pect xe Ye ve fas |S | Ghee bot ia S| AE] SR fe | ema acta 1 xe Rif se [3S | se | ice 1 x RUN x & | Gains Fl x Bo [ae fae | Be |§ i x RS |e] 3) 8 : 1 Ne No | No Ye. i Sn bat No | No | NOTRE] U | eviction Xo | Ne Ne 7 | Sosa a No] Ne | Xo [Be | ts Coss ye no | Ne ne ps Pinetiy-Shapiro’s Yeu No | Yes xe | Gotan mate xs Ro [Re x dv | Siac ats ye [ie Be [Ne x © | Galiatte trongsh ke [xe Ne [ver | sa Le & Gat Be [xe xo fs | MS K | Rroszen’s (4-12 VI- 4-0-3, | Yes | Yes No | No | No [Ney TEE SRO Lvke St Lane taal Das aes aoa Pe OAR) SOM) EMG Mee eek EH Pa: OlMa) © OMA) i Als = Miz (0 0 21) oe Ma = Ma-+ 0; k 8 Property 2 Row and Column scaling fvriance Nort, NoTUnae che menue la rpmmetazed by taking max( ACA, 8), M(B, A)Comparison of Interestingness Measures = Null-(transaction) invariance is crucial for correlation analysis + Lift and x2 are not nullinvariant [SS [Sears Sr] » 5 null-invariant measures Sto) PErgeou acer [lO ol Liftfa, wets wet] No Mik — | No Mik — | Sum (row) wee aes aD bat pe wie [me Tome : Conerence(a, 6) oa] [ves Cosine(a,b) ul [ ves No Coffee |m,~c [~m,~c | ~c ute(a,) to.al] | ves Sum(col.) | m ~m z faxconf(ad)| _max(222ieth 2 oat] \ ves Null-transactions Kulezynski PDE 3 Yatorestingne ure definition nd c measure (1927) variant Data set] me a we Lee ations C aaConp Di 10,000]1,000] 1, 00% }0,000N] 90557] 9.26 [0.91 0.91 [0.91 0.91 Dz___[10,000[1,000] 1,000 \_100 0 L 0.91 0.83 0.91 [o.91| 0.91 Ds 00 [100,000]] 670 [8.44 0.09 0.05. O09 [0.09 0.09 Da 7,000 [1,000] 1, 000\{100,000|/34740/25. 75,8 _ 0.33 05 [05] os—] Ds_({ 1,000] 100 | 10,000 }ia0,000]| 8173 [948 0.09 0.09 a 0.5 0.91 De _N,000[ 10 [100,0087100,000][ 965 [1.97 0.0L 0.7 5 ‘Table 2. Example data Subtle: They disagreeAnalysis of DBLP Coauthor Relationships Recent DB conferences, removing balanced associations, low sup, etc. 1D) Author a Author 6 suplab)[supla)[sup() [Coherence] Cosine | Kule T| Hans-Peter Kriegel Martin Ester 28 | 146 | 54 [| 0.163 (2) [0.315 (7) | 0.355 (9) 2[ Michael Carey Miron Livny 26__| 10d | 58 _[[ 0.191 (1) [0.335 (4) [0.349 (10) 3 | Hans-Peter Kriegel Joerg Sander af 16 | 36 || 0.152 (3) [0.331 (5) | 0.46 (8) 4| Christos Faloutsos [Spiros Papadimitriou] 20 1 26 || 0.119 (7) [0-308 C10)| 0.446 (7) 5 | Hans-Peter Kriegel Martin Pfeifle Cis 6 [0.123 (6) [0.351 (2) | 0.562 27} @ [Hector Garcia-Molina] Wilburt Labio 16_| ad [18 _[P-0.110-9) [oat 8) | 0.500 4 7|_Divyakant Agrawal Wang Hsiung Gig] 2016 533 (5) [0.868 (i [o.se7 a> 8 | Elke Rundensteiner Murali Mani 16 104 2 0.148 10.351 (3) | 0.477 (6) 9 | Divyakant Agrawal Oliver Po rT [iso 100 (10) [0.316 (6) | 0.550 10| Gerhard Weikum Martin Theobald 12 106 iW O.111 (8) [0.312 485 (5) ‘Table 5. Experiment on DBLP data 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 Measures”, Proc. 2007 Int. Conf. Principles and Practice of Knowledge Discovery in Databases (PKDD'07), Sept. 2007 57Which Null-Invariant Measure Is Better? = IR (Imbalance Ratio): measure the imbalance of two itemsets A and B in rule implications |sup(A) — sup(B)| IR(A, B) = —S a ———_ A ) sup(A) + sup(B) — sup(A U B) = Kulczynski and Imbalance Ratio (IR) together present a clear picture for all the three datasets D, through Dg = D, is balanced & neutral = Ds is imbalanced & neutral » Dg is very imbalanced & neutral Data me Tc me me allconf. maa_conf. Kulc. cosine IR Di 70,000 1,000 1,000 T00,000—0.9 O01 TT aor 00 Do 10,000 1.000 1,000 100 0.9L O91 0.9L OL 00 Dg 100 1.000, 1,000 100,000 0.09 0.09 0.09 0.09 00 Da 1,000 1.000 1,000 100,000 05 0.5 0.5 05 (00 Ds 1,000 100 10,000, 100,000, 0.09 0.91 05 0,29 0.89 1.000 10 100.000 100,000 o.01 0.99 05. 0.10 0.99Chapter 5: Mining Frequent Patterns, Association and Correlations: Basic Concepts and Methods @ Basic Concepts m@ Frequent Itemset Mining Methods m@ Which Patterns Are Interesting?—Pattern Evaluation Methods = Summary ©Summary Basic concepts: association rules, support- confident framework, closed and max-patterns Scalable frequent pattern mining methods = Apriori (Candidate generation & test) = Projection-based (FPgrowth, CLOSET+, ...) = Vertical format approach (ECLAT, CHARM, ...) Which patterns are interesting? = Pattern evaluation methodsRef: Basic Concepts of Frequent Pattern Mining (Association Rules) R. Agrawal, T. Imielinski, and A. Swami. Mining association rules between sets of items in large databases. SIGMOD'93 (Max-pattern) R. J. Bayardo. Efficiently mining long patterns from databases. SIGMOD'98 (Closed-pattern) N. Pasquier, Y. Bastide, R. Taouil, and L. Lakhal. Discovering frequent closed itemsets for association rules. ICDT'99 (Sequential pattern) R. Agrawal and R. Srikant. Mining sequential patterns. ICDE'95Ref: Apriori and Its Improvements R. Agrawal and R. Srikant. Fast algorithms for mining association rules. VLDB'94 H. Mannila, H. Toivonen, and A. |. Verkamo. Efficient algorithms for discovering association rules. KDD'94 A. Savasere, E. Omiecinski, and S. Navathe. An efficient algorithm for mining association rules in large databases. VLDB'95 J. S. Park, M. S. Chen, and P. S. Yu. An effective hash-based algorithm for mining association rules. SIGMOD'9S H. Toivonen. Sampling large databases for association rules. VLDB'96 S. Brin, R. Motwani, J. D. Ullman, and S. Tsur. Dynamic itemset counting and implication rules for market basket analysis. SIGMOD'97 S. Sarawagi, S. Thomas, and R. Agrawal. Integrating association rule mining with relational database systems: Alternatives and implications. SIGMOD'98Ref: Depth-First, Projection-Based FP Mining = R. Agarwal, C. Aggarwal, and V. V. V. Prasad. A tree projection algorithm for generation of frequent itemsets. J. Parallel and Distributed Computing, 2002. = G.Grahne and J. Zhu, Efficiently Using Prefix-Trees in Mining Frequent Itemsets, Proc. FIMI'03 = B. Goethals and M. Zaki. An introduction to workshop on frequent itemset mining implementations. Proc. ICDM‘03 Int. Workshop on Frequent Itemset Mining Implementations (FIMI’03), Melbourne, FL, Nov. 2003 = J. Han, J. Pei, and Y. Yin, Mining frequent patterns without candidate generation. sIGMoD’ 00 = J. Liu, Y. Pan, K. Wang, and J. Han. Mining Frequent Item Sets by Opportunistic Projection. KDD'02 = J. Han, J. Wang, Y. Lu, and P. Tzvetkov. Mining Top-K Frequent Closed Patterns without Minimum Support. ICDM'02 = J. Wang, J. Han, and J. Pei. CLOSET+: Searching for the Best Strategies for Mining Frequent Closed Itemsets. KDD'03Ref: Vertical Format and Row Enumeration Methods M. J. Zaki, S. Parthasarathy, M. Ogihara, and W. Li. Parallel algorithm for discovery of association rules. DAMI:97. M. J. Zaki and C. J. Hsiao. CHARM: An Efficient Algorithm for Closed Itemset Mining, SDM'02. C. Bucila, J. Gehrke, D. Kifer, and W. White. DualMiner: A Dual-Pruning Algorithm for Itemsets with Constraints. KDD’02. F. Pan, G. Cong, A. K.H. Tung, J. Yang, and M. Zaki , CARPENTER: Finding Closed Patterns in Long Biological Datasets. KDD'03. H. Liu, J. Han, D. Xin, and Z. Shao, Mining Interesting Patterns from Very High Dimensional Data: A Top-Down Row Enumeration Approach, SDM'06.Ref: Mining Correlations and Interesting Rules S. Brin, R. Motwani, and C. Silverstein. Beyond market basket: Generalizing association rules to correlations. SIGMOD'97. M. Klemettinen, H. Mannila, P. Ronkainen, H. Toivonen, and A. |. Verkamo. Finding interesting rules from large sets of discovered association rules. CIKM'94. R. J. Hilderman and H. J. Hamilton. Knowledge Discovery and Measures of Interest. Kluwer Academic, 2001. C. Silverstein, S. Brin, R. Motwani, and J. Ullman. Scalable techniques for mining causal structures. VLDB'98. P.-N. Tan, V. Kumar, and J. Srivastava. Selecting the Right Interestingness Measure for Association Patterns. KDD'02. £. Omiecinski. Alternative Interest Measures for Mining Associations. TKDE’03. T. Wu, Y. Chen, and J. Han, “Re-Examination of Interestingness Measures in Pattern Mining: A Unified Framework", Data Mining and Knowledge Discovery, 21(3):371- 397, 2010 65