ASSOCIATION RULES & THE APRIORI ALGORITHM BY: JOE CASABONA
INTRODUCTION a. Recap a. Data Mining b. Three types b. Association Rules c. Apriori Algorithm
ASSOCIATION RULES a. Most apparent form of Data Mining b. Objective: Find all co-occurrence relationships among data items c. Strength: Support & Confidence
SUPPORT a. Those who buy X buy Y, where X and Y are sets a. X => Y b. . count = number of occurences c. n = number of total transactions d. Number produced is % of all transactions (T)
CONFIDENCE a. % of transactions where X also contains Y b. Determines predictability of the rule c. Min Support and Confidence Determined.
EXAMPLE a. AR 1: Xbox ---> Controller a. Support: 5/8 b. Confidence: 3/5 b. AR 2: COD 4 ---> Xbox a. Support: 5/8 b. Confidence: 2/5 c. AR 1 passes, AR 2 fails
APRIORI ALGORITHM a. Generate all frequent item sets a. All item sets with min support b. Generate all confident ARs from frequent item sets c. Downward Closure Property
GENERATE FREQUENT ITEM SETS a. Count supports of each individual item b. Create a set F with all individual items with min support c. Creates "Candidate Set" C[k] based on F[k-1]. d. Check each element c in C[k] to see if it meets min support e. Return set of all frequent item sets.
GENERATE CANDIDATE SETS a. Create two sets differing only in the last element, based on some seed set b. Join those item sets into c c. Compare each subset s of c to F[k-1]- if s is not in F[k-1], delete it. d. Return final candidate set
![RULE GENERATE a. Take Frequent Item Set F a. If {F[1], F[2], . .](https://present5.com/presentation/77ab681789d0677a20eb6dfba5ffd84b/image-10.jpg "RULE GENERATE a. Take Frequent Item Set F a. If {F[1], F[2], . .")
RULE GENERATE a. Take Frequent Item Set F a. If {F[1], F[2], . . . F[k-1]} => {F[k]}meets some min confidence, make it a rule b. Remove last element from antecedent, insert into consequent, check again
OTHER ALGORITHMS a. Eclat algorithm b. FP-Growth algorithm c. One-attribute-rule d. Zero-attribute-rule
SAMPLE DATA a. Xbox, Controller, COD 4 b. Xbox, COD 4 c. Xbox, Controller d. Controller, COD 4 e. Xbox, Rock Band, Controller f. Xbox, PS 3 g. COD 4, COD 5, Rock Band h. COD 4, Rock Band • Min Support: 60% • Min Confidence: 50%
RERERENCES The Book I am using: Liu, Bing. Web Data Mining, Chapter 2: Association Rules and Sequential Patterns. Springer, December, 2006 Wikipedia: "Apriori Algorithm. " http: //en. wikipedia. org/wiki/Apriori_algorithm March 23, 2009 "Association rule learning. " http: //en. wikipedia. org/wiki/Association_rules March 25, 2009
Скачать презентацию ASSOCIATION RULES THE APRIORI ALGORITHM BY JOE
77ab681789d0677a20eb6dfba5ffd84b.ppt