Mining Fuzzy Association Rules in a Bank-Account Database by Wai Ho Au and Keith C C Chan Book

Mining Fuzzy Association Rules in a Bank-Account Database by Wai Ho Au and Keith C C Chan Book
  • Page 1

    238IEEE TRANSACTIONS ON FUZZY SYSTEMS, VOL. 11, NO. 2, APRIL 2003Mining Fuzzy Association Rules in aBank-Account DatabaseWai-Ho Au and Keith C. C. ChanAbstract—This paper describes how we applied a fuzzy tech-nique to a data-mining task involving a large database that wasprovided by an internat...

  • Page 2

    AU AND CHAN: MINING FUZZY ASSOCIATION RULES239which describes a person who is single, aged between 35 and 45and with an account balance that is between $1 000 and $2 500,as someone who is likely to use a loan that is between $10 000and $15 000. An association rule defined over market basketdata h...

  • Page 3

    240IEEE TRANSACTIONS ON FUZZY SYSTEMS, VOL. 11, NO. 2, APRIL 2003we describe the bank-account database that was provided by thebank. We then introduce a formalism to handle the union of rela-tional and transactional data in Section IV. The details of FARMII are given in Section V. In this same se...

  • Page 4

    AU AND CHAN: MINING FUZZY ASSOCIATION RULES241Fig. 1.Schema of the bank-account database.TABLE ISUMMARY OF THEBANK-ACCOUNT DATABASEin the bank-account database. A transaction record, therefore,has to store the account involved in the transaction, the date ofthe transaction, the amount of the tran...

  • Page 5

    242IEEE TRANSACTIONS ON FUZZY SYSTEMS, VOL. 11, NO. 2, APRIL 2003is,, where,, can be any attribute in,,or,, or any transformed attribute. In other wordsInstead of performing data mining on the originaland,weperform data mining on.Given a database, different kinds of transformation functionscan be...

  • Page 6

    AU AND CHAN: MINING FUZZY ASSOCIATION RULES243In the case where the value of an attributeof a tuple isunknown, the discretization function, , produces an unknownvalue as its output.The boundaries of the intervals can be specified by usersor determined automatically by using various algorithms (e....

  • Page 7

    244IEEE TRANSACTIONS ON FUZZY SYSTEMS, VOL. 11, NO. 2, APRIL 2003For any categorical attribute,, letdenote the domain of.is repre-sented by linguistic variablewhose value is a linguistic terminwhereis a linguisticterm characterized by a fuzzy set,, so that(2)where. The degree of compatibility ofw...

  • Page 8

    AU AND CHAN: MINING FUZZY ASSOCIATION RULES245If(i.e., the 95th percentile of thenormal distribution), we can conclude that the discrepancy be-tweenandis significantly dif-ferent and, hence,is interesting. Specifically, if thiscondition is satisfied, the presence ofimplies the presenceof. In othe...

  • Page 9

    246IEEE TRANSACTIONS ON FUZZY SYSTEMS, VOL. 11, NO. 2, APRIL 2003. Given the rules that imply the assignment of,, for all,the confidence provided byfor such an assignment is givenby(13)Suppose that, among theattribute values excluding, only some combinations of them,,where, are found to matchone ...

  • Page 10

    AU AND CHAN: MINING FUZZY ASSOCIATION RULES247where Relationship Length is produced by an arithmetic func-tion Relationship Length which is defined as follows:whereis the PROJECT operation in relational algebra andSYSDATE returns the current date in Oracle.This rule states that a customer who has...

  • Page 11

    248IEEE TRANSACTIONS ON FUZZY SYSTEMS, VOL. 11, NO. 2, APRIL 2003algorithms, which discover association rules based on supportand confidence measures, FARM II employs an objective inter-estingness measure to identify interesting associations betweenlinguistic terms without using any user-supplied...