相关测度与增量式支持度和信任度的计算

Correlativity Measure and Incremental Computation of Support and Confidence

通过相关测度的定义,从理论上探讨了增量式规则发现问题,并把分类规则挖掘和关联规则挖掘联系起来进行研究,为该问题的深入研究奠定了理论基础.相关测度刻画了给定关系和相关集合的数字特征.对相关测度的概念、定义、性质以及与支持度和信任度的关系等方面作了详细的分析和探讨,给出了基于相关集合的支持度和信任度的定义及计算方法.证明了测度增量定理和支持度增量定理,并给出了增量式支持度和信任度的计算公式.另外还详细地分析了数据增量对关联规则和信任度的影响,探讨了基于新支持度的候选项的修剪问题.所提出的相关测度及其思想为研究既能用于分类规则又能用于关联规则的统一数据挖掘方法提供了有价值的新思路.

By defining the correlativity measure, the problem of incremental discovering association rule is discussed in theory, and the mining association rule and the mining classification rule are combined to research, which establishes the theoretical foundations for researching the problem in detail. The correlativity measure depicts the numeral character of given relation and mutuality set. The conception, the definition and the properties of the proposed correlativity measure, and the relation between support and confidence are analyzed and discussed in detail. The new definition, methods of computing support, and the confidence based on mutuality set are proposed. The incremental computing formulas of support and confidence are given, and incremental theorems of support and confidence are also proved. On the side, the influences of incremental data upon association rules and the confidence are analyzed in detail. The problem of pruning candidate frequent item set based on new support is also discussed. The correlativity measure and its idea proposed in this paper provide a new valuable way for studying a unification method for mining classification rules and associate rules from database.