E-Retail Example--Model Building

A Web-Mining Scenario Using CRISP-DM

Improved recommendations. Clusterings are produced for varying levels of data integration, starting with just the purchase database and then including related customer and session information. For each level of integration, clusterings are produced under varying parameter settings for the two-step and Kohonen network algorithms. For each of these clusterings, a few C5.0 rulesets are generated with different parameter settings.

Improved site navigation. The Sequence modeling node is used to generate customer paths. The algorithm allows the specification of a minimum support criterion, which is useful for focusing on the most common customer paths. Various settings for the parameters are tried.