29. October 2015

Enhanced Methods for Data Visualization

The visualization of data and results in charts is a common method to make information transparent.

You often find charts like line charts (temporal processes), scatter charts (showing the relation between two variables), histograms (distribution) and box-whisker charts (distribution and outlier).

A rather uncommon chart type is the parallel-coordinate chart. All variables are shown on the x-axis. Each data point is then visualized as a line. Moreover, the lines are colored according to the progression of a variable so that similar data points can easily be identified.

This chart type is often used for the visualization of multicriteria optimization results. The example above shows the optimization of an inlet port concerning tumble and mass flow. The eight values on the left describe the geometry of the inlet port. Each line corresponds to an optimal design with its characteristics, i.e., parameter settings. The lines are colored according to the flow rate (red = low, blue = high).

Now it is easy to see that the objectives tumble and mass flow are working in opposite directions. Furthermore, the charts shows the relation between the cross section and the objectives (small cross section = low flow rate, high cross section = high flow rate).

Consequently, a parallel-coordinate chart is also suitable for a comparison of high dimensional data points.