Analyzing data

After all simulations are finished you may execute RAWSimO.DataPreparation.exe on your machine, choose "Process experiment results" and input the path to the directory in which all result directories are. The tool will then try to generate some graphs and also combine all results in one .csv file (footprints.csv) that can be passed to Excel or similar. For the graph generation a working installation of GnuPlot is necessary (see http://www.gnuplot.info/).

You can use the "footprints.csv" exported by RAWSimO.DataPreparation.exe (see above) to analyze the behavior of different mechanisms on different layouts and settings. The different values should be somewhat self-explanatory by their name. If not, you can lookup a short description in: RAWSimO.Core.Statistics.FootprintDatapoint.FootPrintEntry and the way they are calculated in the constructor of that class (FootprintDatapoint). Furthermore, you can use the graphs plotted by "Process experiment results" (see above) to investigate single runs. Doing this you can look at more details in a time-based way of the particular run. This may help explaining counter intuitive behavior of methods. The GnuPlot scripts are still contained in the directories and can be modified to better suite your needs.

See a sample of a time-based plot below:

It is also possible to generate scatter plots of all runs of an experiment. For this execute RAWSimO.DataPreparation and choose "Scatterplot experiment results". This generates a set of differently colored plots (by instance and mechanisms) for different output values in comparison.

See a sample of a scatter plot below: