The CRISP-DM methodology considers the analytical activities as a cyclical set of phases to be repeated until a satisfactory result is obtained. Not surprisingly then, Crisp-DM methodology phases are usually represented as a circle going from business understanding to the final deployment:
As we can see within the diagram, the cycle is composed of six phases:
- Business understanding
- Data understanding
- Data preparation
- Modeling
- Evaluation
- Deployment
This is the greater abstraction level of the Crisp-DM methodology, meaning one that can apply, with no exception, to all data mining problems. Three more specific layers are then conceived as a conjunction between the general model and the specific data mining project:
- Generic tasks
- Specialized tasks
- Process instances
All of the components of every level are mapped to one component of the layer above, so that when dealing with a specific data mining problem, both bottom-up and top-down approaches are allowed, as we will see in the last paragraph of this chapter.