STAGES OF SOM-BASED EXPLORATORY DATA
ANALYSIS

 

In this section guidelines are presented for using the SOM in practical applications such as the case studies in the Publications belonging to this thesis. Related treatments have been presented earlier [Kaski and Kohonen, 1995, Kohonen, 1995c].

The basic steps in applying the SOM are quite standard although, as usual in exploratory data analysis, care must be taken in choosing the inputs and in evaluating the results to avoid misleading conclusions. The steps of the analysis can be used interactively by modifying the composition of the data set on the basis of insight gained during the previous round.

The first step in the analysis consists of the choice of the data set; although this step seems evident it cannot be stressed too much. No matter how good the methodology is the results will fundamentally depend on the quality and suitability of the data. Even good methods may only aid in the process of gaining insight into the nature of the data.

• Preprocessing
• Computation of the maps
• Choosing good maps
  • Note 1: Sparse data.
  • Note 2: High dimensionality.
  • Note 3: Computational complexity.
• Interpretation, evaluation, and use of the maps
  • Interpretation.
  • Evaluation.
  • Use of the organized maps for exploratory data analysis.