Download
1 / 39
420 likes | 767 Vues
Mining Association Rules. Charis Ermopoulos Qian Yang Yong Yang Hengzhi Zhong. Outline. Basic concepts and road map Scalable frequent pattern mining methods Association rules generation Research Problems. Frequent Pattern Analysis.
E N D
Mining Association Rules Charis Ermopoulos Qian Yang Yong Yang Hengzhi Zhong
Outline • Basic concepts and road map • Scalable frequent pattern mining methods • Association rules generation • Research Problems
Frequent Pattern Analysis • Frequent pattern: a pattern (a set of items, subsequences, substructures, etc.) that occurs frequently in a data set • Frequent pattern mining • Finding inherent regularities in data • Foundations of many data mining tasks (association, correlation, classification etc.) • Applications: Basket data analysis, cross marketing, catalog design, web log analysis.
What is association rule mining ? • Given: (1) database of transactions, (2) each transaction is a list of items (purchased by a customer in a visit) • Find: all rules that correlate the presence of one set of items with that of another set of items • E.g., 98% of people who purchase tires and auto accessories also get automotive services done • Itemset X = {x1, …, xk} • Find all the rules X Ywith minimum support and confidence
What is association rule mining ? • Support = The rule X => Y has supports in the transaction set D if s% of transaction in D contain X union Y • Confidence = The rule X => Y has confidence c if c% of transactions in D that contain X also contain Y • Association rule mining • Find all sets of items that meet the minimum support - Frequent pattern mining • Generate association rules from these sets
Generate Frequent Itemsets • 3 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 • Initially, scan DB once to get frequent 1-itemset • Generate length (k+1) candidate itemsets from length k frequent itemsets • Testthe candidates against DB • Terminate when no frequent or candidate set can be generated
Supmin = 2 L1 C1 1st scan C2 C2 L2 2nd scan L3 C3 3rd scan
Apriori candidate generation • 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}
Drawbacks • Multiple scans of transaction database • Multiple database scans are costly • Huge number of candidates • To find frequent itemset i1i2…i100 • # of scans: 100 • # of Candidates: (1001) + (1002) + … + (110000) = 2100-1 = 1.27*1030 !
FPGrowth • Uses the Apriori Pruning Principle • Scan DB only twice! • Once to find frequent 1-itemset (single item pattern) • Once to construct FP-tree, the data structure of FPGrowth
FPGrowth TID Items bought 100 {f, a, c, d, g, i, m, p} 200 {a, b, c, f, l, m, o} 300 {b, f, h, j, o, w} 400 {b, c, k, s, p} 500{a, f, c, e, l, p, m, n} Header Table Item frequency f 4 c 4 a 3 b 3 m 3 p 3 TID (ordered) frequent items 100 {f, c, a, m, p} 200 {f, c, a, b, m} 300 {f, b} 400 {c, b, p} 500 {f, c, a, m, p} {} f:1 c:1 a:1 m:1 p:1
FPGrowth TID Items bought 100 {f, a, c, d, g, i, m, p} 200 {a, b, c, f, l, m, o} 300 {b, f, h, j, o, w} 400 {b, c, k, s, p} 500{a, f, c, e, l, p, m, n} Header Table Item frequency f 4 c 4 a 3 b 3 m 3 p 3 TID (ordered) frequent items 100 {f, c, a, m, p} 200 {f, c, a, b, m} 300 {f, b} 400 {c, b, p} 500 {f, c, a, m, p} {} f:4 c:1 c:3 b:1 b:1 a:3 p:1 m:2 b:1 p:2 m:1
{} Header Table Item frequency head f 4 c 4 a 3 b 3 m 3 p 3 f:4 c:1 c:3 b:1 b:1 a:3 p:1 m:2 b:1 p:2 m:1 FPGrowth TID (ordered) frequent items 100 {f, c, a, m, p} 200 {f, c, a, b, m} 300 {f, b} 400 {c, b, p} 500 {f, c, a, m, p}
{} Header Table Item frequency head f 4 c 4 a 3 b 3 m 3 p 3 f:4 c:1 c:3 b:1 b:1 a:3 p:1 m:2 b:1 p:2 m:1 FPGrowth Conditional pattern bases Item cond. pattern base freq. itemset p fcam:2, cb:1 fp, cp, ap, mp, pfc, pfa, pfm, pfm, pca, pcm, pam, pfcam m fca:2, fcab:1 fm, cm, am, fcm, fam, cam, fcam b fca:1, f:1, c:1 … a fc:3 … c f:3 …
FPGrowth vs Apriori • 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
FPGrowth vs Apriori Data set T25I20D10K
FPGrowth vs Apriori • Dense dataset (http://www.cs.yorku.ca/course_archive/2005-06/F/6412/lecnotes/assorule3-2.pdf)
Scaling • DB projection • FP-tree cannot fit in memory? • Partition a database into a set of projected DBs • Construct and mine FP-tree for each projected DB
Tran. DB fcamp fcabm fb cbp fcamp p-proj DB fcam cb fcam m-proj DB fcab fca fca b-proj DB f cb … a-proj DB fc … c-proj DB f … f-proj DB … am-proj DB fc fc fc cm-proj DB f f f … DB Projection
Charm: Closed Itemset Mining • A frequent itemset with size s has 2s-2 frequent subsets • S is large in many real world problems • biosequences, census data, etc • Only generate itemsets that cannot be subsumed others with the same support • If A B, and sup(A) = sup (B), A will not be in the result • But sup(A) can be inferred from B and others
Closed Itemset • An itemset X is closed if X is frequent and X does not have a superset Y such that supt(Y) = support(X) • Lossless compression • Divide frequent itemsets into equivalence classes
Charm: Search in IT-Tree • Each node has a Itemset and its Tidset
Charm properties to prune • Itemset Xi, Xj, and theri Tidset t(Xi), t(Xj) • If t(Xi) = t(Xj), sup(Xi) = sup(Xj) = sup(Xi Xj) • Xi, Xj always occurs together • If t(Xi) t(Xj), sup(Xj) != sup(Xi) = sup(Xi Xj) • If Xi occurs, Xj also occurs • If t(Xj) t(Xi), sup(Xi) != sup(Xj) = sup(Xi Xj) • If Xj occurs, Xi also occurs • If t(Xj)!= t(Xi), sup(Xi) != sup(Xj) != sup(Xi Xj)
T x 1356 A x 1345 D x 2456 C x 123456 W x 12345 TA x 135 TW x 135 DT x 56 DW x 245 DA x 45 Minimum support = 3 WC x 12345 AW X 1345 AWC x 1345 TC x 1356 DC x 2456 TAC x 135 TWC x 135 DWC X 245 TAWC x 135
Tids Itemsets Charm: diffset for fast couting • Maintain a disk-based tidset for each item • Vertical • Easy to compute support: cardinality • Intersection is expensive when tidset is large • Diffset • Track the differences in tids of a child node from its parent • Save memory when tidsets are large and differences are little X d(XY) Y
Association Rules AB …A implies B The easiest way to mine for association rules, is to first mine for frequent itemsets.
Mining Association Rules • Quantity Problem • Many frequent itemsets, many rules • Redundant rules • Quality Problem • Not all rules are “interesting”
Association Rules • Frequent itemset : AB • Derived rules : AB and BA Support (AB) = P(A,B)=|AB|/N Confidence (AB) = P(B|A)=P(B,A)/P(A)=|AB|/|A| |AB|: count of (AB) |A|: count of (A) |B|: count of (B) N: total number of records
Other measures of Interestingness • 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
Various Kinds of Association Rules • multi-level association Data categorized in a hierarchy • multi-dimensional association age(X,”19-25”) occupation(X,“student”) buys(X, “coke”) • quantitative association Rules with numerical attributes • “interesting” correlation patterns Eg. Some items (e.g. diamonds) may occur rarely but are valuable
Constraint-based 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 • System optimization: explores such constraints for efficient mining—constraint-based mining
Constraints in Data Mining • Data constraint— using SQL-like queries • find product pairs sold together in stores in Chicago in Dec.’02 • Dimension/level constraint • in relevance to region, price, brand, customer category • Rule (or pattern) constraint • small sales (price < $10) • Interestingness constraint • strong rules: min_support 3%, min_confidence 60%
Anti-Monotonicity in Constraint Pushing • Anti-monotonicity • When an intemset S violates the constraint, so does any of its superset • sum(S.Price) v is anti-monotone • sum(S.Price) v is not anti-monotone • Example. C: range(S.profit) 15 is anti-monotone • Itemset ab violates C • So does every superset of ab
Monotonicity for Constraint Pushing • Monotonicity • When an intemset S satisfies the constraint, so does any of its superset • sum(S.Price) v is monotone • min(S.Price) v is monotone • Example. C: range(S.profit) 15 • Itemset ab satisfies C • So does every superset of ab
Database D L1 C1 Scan D C2 C2 L2 Scan D L3 C3 Scan D Constraint: Sum{S.price} < 5 The Constrained Apriori Algorithm: Push an Anti-monotone Constraint Deep
Converting “Tough” Constraints • Convert tough constraints into anti-monotone or monotone by properly ordering items • Examine C: avg(S.profit) 25 • Order items in value-descending order • <a, f, g, d, b, h, c, e> • If an itemset afb violates C • So does afbh, afb* • It becomes anti-monotone!
