Show TOC

Process documentationClustering

 

Clustering is used to segment data automatically and to split the data into clusters. The system determines previously unknown groups of related data within an unstructured dataset. At the same time, both the segmentation criteria, in other words the clusters themselves, as well as the assigned data records are determined.

You perform the segmentation by training a model based on historical data. You can use this segmentation via a forecast on another dataset.

For more information, see Clustering.

Process

This process runs in SAP NetWeaver BW:

  1. Create clustering model

    You create a clustering model with the relevant dimensions and weighting. The dimensions are the input criteria for the clustering. On a technical level, they are the model fields of a clustering model.

    If you do not know which dimensions need to be considered in clustering, you can use the decision tree of the data mining method to determine the most important dimensions from a larger amount of possible dimensions and then apply them in clustering. For more information, see Data Mining.

  2. Train and execute clustering model

    While training the model, the system determines the customer segments in SAP NetWeaver BW. To train the model, you use a BW query that provides the data on your customers that is collected in the customer knowledge base. The system groups customers together by shared properties and creates clusters – and thereby customer segments. Each customer is assigned to exactly one customer segment.

    For information on the extractors required, see:

    • Generic Extractor for Marketing Attribute Values

    • Generic Extractor for Marketing Attribute Values - Texts

  3. Transfer results of clustering model

    There are different ways with which you can transfer the clustering results from SAP NetWeaver BW into SAP CRM. For this, you can use the Analysis Process Designer to transfer the master data attribute updated in SAP NetWeaver BW to a marketing attribute.