Efficient Data Mining Technique Using Associate Rule

Abstract

In recent years there has been a massive proliferation of data that is accessible electronically, ranging in form from highly structured
databases, such as article collections and company records, to web sites and pages that vary greatly in content. Since its introduction, association
rules mining technique became one of the most frequently used data-mining techniques. Association rules have exhibited an excellent ability to
identify interesting association relationships among a set of binary variables describing huge amount of transactions. Although the technique can
be relatively easily generalized to other variable types, the generalization can result in a computationally expensive algorithm generating a
prohibitive number of redundant rules of little significance. This danger especially applies to quantitative (and ordinal) variables. These issues
are tackled in this thesis and an alternative approach to the quantitative association rule mining is presented and verified. In the proposed
approach, quantitative or ordinal variables are not immediately transformed into a set of binary variables. Instead, simple arithmetic operations
are applied in order to construct the compound attributes and the algorithm further searches for areas of increased association which are finally
decomposed into conjunctions of atomic attributes.

 

 

Keywords- Data Mining, Association Rule and Efficient Search Engine.