Association rule
Association rule/correlation rule

When a certain phenomenon occurs, it seems that, another phenomenon occurs, it is something related to between the strong phenomenon combination of the phenomenon whose simultaneity and related characteristic are strong, or such as that.

The technology which extracts association rule from the mass data which is accumulated to the database is called association rule extractionorassociation analysis.It is utilized in marketbasket analysis.

With association analysis,bythe fact that the mass data which was collected with the POS system and the like is processed on the basis of fixed algorithm, "as for 90% of the simultaneous buyers of the bread and the butter the kind of conclusion which purchases also the milk", is deduced. Such item (the commodity) combination with association rule, loads association analysis function with many data mining tools,

[ Bread and butter ] ? [ milk ] (degree of support 4%, reliability 90%)

As with said, it is expressed. This time, bread and butter conditional section (antecedent: Also prerequisite section), milk conclusion section (consequent: Also the conclusion section is called) with. Both the conditional section and the conclusion section it is possible to include the plural items.

With association analysis, from ratio of the conditional section and the conclusion section for transaction number of cases reliability (confidence: Also degree of conviction) with degree of support (support: ) It calculates also support, it appraises criticality of rule on the basis of this index (the sectionof association analysis reference).

As for expression of association rule being clear, there is a merit that quite to understand it is easy to cause actual action easily. However, association rule just shows the fact that "correlation (the association) it is", is not something which the causality the truth is proven.

In addition with association analysis, rule and the natural rule which exist everywhere are deduced in large quantities. In addition, expression being clear, is not understanding impossible, if there is many also the fact that the rule which cannot discern the background is guided, does not investigate this with expedient other than the data mining, it cannot judge "the true value" of rule.

Really to verify with the counter and the like it is indispensable, for background understanding of true causal relation and customer conduct, the manager et. al. of the retail store to raise hypothesis even from such point as rule on the basis of the contents which are suggested.