The induction of subgroups, described in Section 2.2, represents the main step of the proposed descriptive induction process. This step corresponds to the data mining step of the standard process of knowledge discovery in databases (KDD). The overall descriptive induction process, proposed in this paper, is comparable to the standard KDD process (Fayyad, Piatetsky-Shapiro, & Smyth, 1996), with some particularities of the task of subgroup discovery.

The proposed expert-guided subgroup discovery process consists of the following steps:

- 1.
- problem understanding
- 2.
- data understanding and preparation
- 3.
- subgroup detection
- 4.
- subgroup subset selection
- 5.
- statistical characterization of subgroups
- 6.
- subgroup visualization
- 7.
- subgroup interpretation
- 8.
- subgroup evaluation

Section 3.1, illustrating steps 1 and 2, presents a medical problem used as a case study for applying the proposed descriptive induction methodology. Tools for supporting subgroup detection and selection in steps 3 and 4 were described in detail in Sections 2.2 and 2.3, while the results of expert-guided subgroup detection and selection are outlined in Section 3.2. Methods and results of steps 5-8 for this domain are outlined in Sections 3.3-3.6, respectively.

The proposed descriptive induction process is iterative and interactive. It is iterative, since many steps may need to be repeated before a satisfactory solution is found. It is also interactive, assuming expert's involvement in most of the phases of the proposed descriptive induction process. The expert's role in the patient risk group detection application is described in Section 3.7.

- The Problem of Patient Risk Group Detection
- Results of Expert-Guided Subgroup Detection and Selection
- Statistical Characterization of Subgroups
- Subgroup Visualization
- Subgroup Interpretation through Visualization
- Subgroup Evaluation
- The Expert's Role in Subgroup Discovery