Data Mining

Mining Multiple-Level Association Rules in Large Databases

Developing a top-down progressive deepening method

Introduction

When mining multiple levels of data from large data sets , an efficient set of association rules should be in place . When examining multiple-level association rule mining , there must be multiple levels of extraction and efficient methods for multiple-level rule mining and a top down deepening method based on the Apriori principle . This principle holds that if the initial item set is frequent then all of its subsets are frequent therefore frequent data items are at the top

most level and their descendants are at lower concept levels . Efficient level shared mining and algorithms can be explored and developed from large databases using a progressively deepening method based on the Apriori principle

To make sure that there are multiple levels of association it is necessary to provide and store concept classifications ranging from basic to high level . For an efficient method of multiple-level rule mining there are a couple of approaches to consider . One is a direct application of single level mining rule association to multiple level mining i .e . the same minimum support and minimum confidence thresholds While this is excellent at high level abstractions it can produce weak associations at lower levels

The second approach is to apply different minimum support thresholds and different confidence thresholds at varying levels of abstraction . By progressively reducing the minimum support thresholds at lower levels of abstraction rules can be mined at multiple concept levels thereby discovering other association...