How to demonstrate the efficiency of a dynamic factor in comparison with some constant coefficient with a chart?

Question

I proposed a measure which estimates the coefficient of each sample dynamically, So the coefficient of two sample s1 and s2 can be different. To show the effectiveness of this measure, I want to compare it with different constant coefficients. The constant coefficients are in the range of 0.1 to 0.6. When the constant coefficient Y is for example 0.2, it is equal for all samples. There are 1000 samples in the dataset.

I want to represent the effectiveness in a chart.

The scores are as follows:

• The score of dynamic measure: 0.9

• The score of constant coefficients Y:

  • Y=0.1 ⇒ score=0.5
  • Y=0.2 ⇒ score=0.6
  • Y=0.3 ⇒ score=0.65
  • Y=0.4 ⇒ score=0.55
  • Y=0.5 ⇒ score=0.45
  • Y=0.6 ⇒ score=0.40

How can I show the differences between the effectiveness of dynamic measures in a chart that clearer than what I already have?

I used the following chart, but it is not clear for some of the readers. So, I would like to make some changes to make it more readable.

Any help is appreciated.

Answer 1

Obviously the efficiency is one number which is calculated from a time series of unpublished system output variables and the cumulated cost during the run. The input variables of the system are set by a control system. You have done also six test runs with forced input variable value sets which were constant during the test runs. Those input variable value sets are named as "Constant" is 0,1 x K where K is the number of the tst run. You do not want to publish how many input variables there is and what they are.

I believe the system controller has a set of tuning parameters which finally cause how the closed loop system performs. Your efficiencies are a mapping from tuning parameter vectors to real numbers between 0 and 1. That mapping can be presented as discrete points in a space which has 1 dimension more then there's tuning parameters for the system controller If only one tuning parameter has been changed between measuring runs, then you can plot points on a plane and write a list of fixed tuning parameters. If there's 2 tuning parameters which are changed between the measuring runs, you can still stay in plane and indicate the second tuning parameter values with colors or plot symbols. As well you can have a plot in 3D space (X and Y are the tuning parameters during the run, Z=the calculated efficiency after a run) If you are lucky there seems to be a curve in one tuning parameter caseand in 2 tuning parameters case there seems to be a family of curves in a plane or a surface in 3D space.

You simply draw the six forced run efficiencies with a line in a 2D plane or a plane in 3D space. If some special tuning parameter combinations happen to cause those 6 discrete constant system input variable combinations named C=0.1 , 0.2 , 0.3 , 0.4 , 0.5 or 0.6 , you of course exaggerate with an auxiliary symbol those tuning parameter combinations which would result the six constant input efficiencies.