Contents
General Selection:
The selection of data mining and visual analytics procedures requires a certain amount of experience as well as several iterations of different procedures and visualizations.
Therefore, this section provides a rough introduction to the work flow of a typical simulation data analysis.
For the flow of the data mining and visual analytics methods, the flowchart in Figure 1 can be used as a guide.
Figure 1 - Data mining and visual analytics flowchart.
The first task is to identify interesting result parameters.
These are result parameters that have either statistical anomalies or contextual relevance to the task at hand.
To find statistical peculiarities, the correlation analysis is suitable for a first impression over all parameters.
Histogram or scatterplot analysis can be used to investigate 1- or 2-dimensional relationships.
If different scenarios or strong structural differences have been defined for the simulation model, these should be considered as a whole but also filtered.
Once the conspicuous result parameters have been identified, the clustering method can be carried out on their basis.
With the results of the clustering (assigned class) the investigation of the relations between result, input and class data can be started.
One of the following methods is suitable for this purpose:
•Radar plots
•Regression analysis
•Classification trees
•Scatterplot analysis
•Parallel plots
Through these methods, knowledge about the parameters and relationships can be gained, which in turn can be used to draw conclusions about the system.
If no insights are generated by the current parameters and settings, this process is iterative.
For the handling and the meaningful use of the methods a certain measure of experience and understanding is needed.
Therefore, the following sections provide a brief introduction to the use and interpretation of the methods.
© SimPlan AG - Hanau District Court, Commercial Register (Part B) 6845 - info@simplan.de - www.simplan.de/en