Affinity Analysis 
This InfoArea contains objects that are required within the context of affinity analysis.
In affinity analysis, you investigate your sales data for correlations between the goods sold. The result of the affinity analysis is a set of rules of form
"Item B follows from Item A".
These rules mean that the sale of Item A (also called premiss) causes to a "not insignificant degree" also the sale of Item B (also called conclusion). You can define which rules are to be considered 'interesting'. You can define the term 'Items' in different ways.
You can define single articles as an item, or each grouping of articles that is represented by a single InfoObject, for example, material group, subcategory, customer-specific objects.
Depending on the choice of item the rules are very specific but are sometimes too numerous (when considering single articles) or very general, with a small number of rules . The choice of the appropriate item level is very much dependent on the assortment structure in each case.
The information on product affinities is useful for business decisions, for example, the arrangement of goods in the store or the choice of appropriate sales offers or bonus buy schemes.
The following diagram shows the most important components of affinity analysis and the data flow between these components.
Note
Note: The use of the SAP NetWeaver BI Accelerator (BIA) with minimum revision 7.00.44.00. is an absolute prerequisite for use of affinity analysis.
Affinity analysis makes use of the functionality of the BIA association analysis. The system requests only rules that show exactly one item for the premiss and conclusion of the rule.
For more information on BIA association analysis, see SAP NetWeaver BI Accelerator.
The BIA index expects the data to be analyzed in an InfoCube. The system loads the data (for example, from the POS Inbound Processing Engine (PIPE) into this InfoCube and replicates the data then using the standard tools of the BI in the corresponding BIA index. Th e call-up and the receipt of the analysis results takes place by way of a separate transaction (RSBCT_RTM01). You can save the results of the affinity analysis directly in a real-time InfoCube. There the results are available for further processing (queries, update to further data targets).
On the BI side, two things are important for affinity analysis: The analysis configuration and the analysis run.
The analysis configuration is a Customizing setting in which you can specify basic aspects of the affinity analysis. These include:
the InfoCube with the TLOG data to be analyzed
which characteristic is to be seen as an item
An analysis run represents a call-up of the affinity analysis including the related parameters and found rules. An analysis run refers to exactly one analysis configuration and possesses the following attributes:
Selection options
Requested technical key figures for the rules
Features
Within an analysis run, you can determine multiple groups of rules. Each group of rules is valid for a different subset of the transaction data. Here you make settings in the relevant analysis configuration showing which characteristics of the TLOG Cubes are to be regarded as rule keys. For each combination of the key characteristics, the system carries out a separate analyis internally.
You can also define characteristics of the TLOG Cubes as filter characteristics. Using these, you restrict the quantity of transaction data without this selection being directly visible in the rules found.
You can use navigation attributes for both the rule key and filter characteristics. BW hierarchies and time-dependent attributes are, however, not supported.
In addition to the selection of filter characteristics, you can restrict the quantity of the item combinations to be searched. You can define both a minimum support for the item combination and also the maximum number of item combinations with maximum support ("Top X Items").
You can, just as with the quantity of items to be considered, reduce the quantity of returned rules:
You can specify a minimum confidence.
You can specify the maximum number of rules with maximum "goodness" in each case ("top X rules"). You can define which key figure is to be considered as the significance measure of a rule.
The following key figures are available as significance measures:
To shorten the runtime of the analysis, you can specify a sampling rate. This way the system analyzes only a portion of the relevant transactions. In this case, the system may possibly not provide the same results as with analyis of all relevant data.
Constraints
Exactly one Item is expected on the right and left of the rules. Rules of form
"Item C follows from Item A and Item B".
are not supported by the system.