Applying Data Mining to Field Quality Watchdog Task Satoshi HORI, Member (Institute of Technologists), Hirokazu TAKI, Member (Wakayama University), Takashi WASHIO, Non-member, Motoda Hiroshi, Non-member (Osaka University) This article describes a watchdog program that discovers meaningful repair cases from a field service database. Meaningful cases are those judged worth probing further to prevent an epidemic of quality problems. Our system has employed the Apriori algorithm, a data mining technique which efficiently performs the basket analysis. Our system proves that this data mining technique is not only useful in knowledge discovery but is also capable of performing the database watchdog task. The Apriori algorithm automatically generates frequent itemsets from a large set of records. A frequent itemset is an arbitrary combination of values that appear more often than a threshold minimum support. The algorithm often generates too many itemsets for quality engineers to review carefully in their daily work. Many itemsets don t provide sufficient information to investigate further. Hence, in order not to generate these valueless itemsets, the Apriori algorithm is modified in two ways. One way is Basket analysis on objective and explanatory attributes, and the other is Itemset reduction. The advantage of our method is demonstrated with some experimental results.,,,,, 1. Apriori alogrithm 11 1 (Relational Database) C121 1 13 1
1... (epidemic quality problem) Excel SQL(Structured Query Langauge) 12 1 (P i,p j ) SQL (P i,p j ) P i P j SQL 2 (3) 1 minimum support() 2 Apriori algorithm (1) Apriori algorithm (Frequent itemset) Apriori algorithm Apriori algorithm 2. (item) (Record/Transaction) (:Frequent itemset) (association rules) Agrawal Apriori Algorithm (1) (SQL) Apriori Algorithm 21 Apriori Algorithm (Record/Transaction) (item) minimum support( min-sup) minimum confidence( min-conf) 1 min-sup {bread, milk, coffee,...}. 2 2 T.IEE Japan, Vol. 121-C, No.1, 101
A B ( A, B, C A B = φ, A B = C) ) C = {bread, milk, butter} bread milk butter min-conf :{bread} {milk, butter} Apriori algorithm 1 Apriori algorithm Step-1. 1 item frequent 1-itemset Step-2. apriori-gen frequent k-1 itemset k-itemset Step-3. k-itemset t 1 Step-4. min-sup k-itemset frequent k-itemset Srikant item constraint (6) (taxonomy) (5) bread, milk,...) 3. 1 Apriori Algorithm 22 (SQL) X ( ) min-sup Apriori algorithm min-sup=5 1479 759 1 Apriori algorithm 2 C121 1 13 3
31 2 ID / Ex. Model-A C103 X ) ID 2 2 2 STEP-1. N=1. STEP-2. N STEP-3. Step-2 N+1 minsup STEP-4. Step-2 L o STEP-5. lo k L o STEP-5.1. lo k = {o 1,...,o m } lo k k o i STEP-5.2. le,1 k,...,lk e,n STEP-5.3. n l k = lo k. i=1 32 3 4 Step-1. 4 Step-2. Apriori algorithm L o Step-3. Step-2 L o Apriori algorithm L Step-4. L l k e,i 4 T.IEE Japan, Vol. 121-C, No.1, 101
1 2 3 4 4 1479 759 STEP- 5 28 ( 11 6 ) 4 (a) 251 414 492 759 28 7 13 16 4 ) min-sup 3 Freq. 1497 (a) Apriori 759 (b) Apriori 28 4 ID (a) (b) 4 Case-A 11 251 7 Case-B 7 414 13 Case-C 6 492 16 3 Watchdog 4. 33 1479 6 ( ( ) min-sup 5 3 (a) 6 Apriori algorithm (b) Apriori algorithm 3 Apriori algorithm L SQL Apriori C121 1 13 5
SQL min-sup Huber (7) 12 1 31 13 99 99 1 R.Agrawal, R.Srikant, Fast algorithms for mining association rules, Proc. of 20th VLDB Conference, pp.487-499 (1994). 2 Y.Cai, N.Cercone, J.Han, Attribute-oriented induction in relational databases, Knowledge Discovery from Databases (Ed. Piatesky-Shapiro) MIT Press, pp.214-228 (1991). 3 U.M.Fayyad, et. al. Ed., Advances in knowledge discovery and data mining (1996). 4 ed., (),, Vol.12, No.4, pp.496-549 (1997). 5,,,, /, pp.103-108 ( 1997). 6 R.Srikant, Q.Vu, R.Agrawal, Mining association rules with item constraints, Proc. of 3rd Int l Conf. on Knowledge Discovery and Data Mining, pp.67-73 (1997). 7 P.J.Huber, From Larget to Huge: A Statistician s Reactions to KDD & DM, Proc. of 3rd Int l Conf. on Knowledge Discovery and Data Mining, pp.304-308 (1997). 1982 3 4 1988 PURDUE 2001 1980 3 1986 1990 1998 4 1988 1990 1996. AAAI 1967. 1975 19771984 19891992, 2001 2000 AAAIIEEE Computer Society 6 T.IEE Japan, Vol. 121-C, No.1, 101
4 C121 1 13 7