Typical tasks of data mining are Outlier Detection, Cluster Analysis, Classification, Association analysis, Regression analysis. These tasks can be roughly divided into observation problems (outlier detection, cluster analysis) and forecasting problems (classification, regression analysis).
Association analysis refers to the search for strong rules. These consequent association rules describe correlations between things that occur together. So, the purpose of an association analysis is to identify items (elements of a set, such as individual items in a shopping cart) that imply the occurrence of other items within a transaction. A relationship between two or more items revealed in this way can then be represented as a rule of the form “If item A, then item B” or A → B.
A typical field of application is the correlations during purchasing, the so-called shopping basket analysis, in order to initiate targeted advertising measures. For example, 80 percent of the purchases where beer is purchased are also potato chips. Both products appear in 10 percent of purchases. These insights are often used in cross-marketing.
---
Let us give an association rule {toothbrush} → {toothpaste}. Support calculates the percentage of all transactions that are subject to the {toothbrush} → {toothpaste} rule. For the calculation, the number of transactions in which both items of interest occur is divided by the number of all transactions.
In what proportion of transactions in which {toothbrush} also appears {toothpaste}? To calculate confidence, the number of all rule-fulfilling transactions is divided by the number of transactions containing {toothbrush}.
Let’s say 10 percent of all customers buy {toothbrush, toothpaste}, 20 percent of all customers buy {toothbrush}, and 40 percent of all customers buy {toothpaste}. Then the rule has a lift of 1.25.
Algorithms shall be designed in such a way that all association rules are found with a predetermined minimum confidence and support. The methods should not require assumptions about the characteristics to be analysed. This would not be conceivable, for example, in a mail-order business with many thousands of items.
