关联规则挖掘中几个兴趣度量的值域研究

Research on the Value Domains of Several Interest Measures in Association Rule Mining

摘要: 本文旨在研究关联规则挖掘中的各种兴趣度量的值域问题。首先，详细介绍了关联规则挖掘过程中涉及的定义和支持度、置信度、确信度、提升度和Laplace测度这五种兴趣度量的定义，并通过具体例子对这些度量进行了说明和解释。然后，深入探讨了这五种兴趣度量的值域，并给出了其在数据库大小有限和接近无穷两种情况下的值域情况。此外，本文还对这些兴趣度量值域的区间端点的取值进行了细致讨论，指出了与其他研究结果的区别及其原因，并给出了严谨的数学证明和对比分析，为关联规则挖掘提供了更全面和准确的度量工具。

Abstract: This article aims to study the value domains problem of several interest metrics in association rule mining. Firstly, a detailed introduction was given to the definitions of five interest measures involved in the process of association rule mining, including support, confidence, conviction, lift, and Laplace measures. These measures were explained and illustrated through specific examples. Then, the value domains of these five interest measures were explored in depth, and their value situations were given in two scenarios: limited database size and near infinite database size. In addition, this article also provides a detailed discussion on the values at the interval endpoints of these interest measures, pointing out the differences and reasons from other research results, and providing a more comprehensive and accurate measurement tool for association rule mining through rigorous mathematical proof and comparative analysis.