Difference between revisions of "Modeling"
Jump to navigation
Jump to search
Line 20: | Line 20: | ||
funcAresults[a] = functionAcall(p1,p2) | funcAresults[a] = functionAcall(p1,p2) | ||
print(funcAresults) | print(funcAresults) | ||
+ | </syntaxhighlight> | ||
+ | * The [https://en.wikipedia.org/wiki/Approximation_error Approximation error] ([https://nl.wikipedia.org/wiki/Benaderingsfout benaderingsfout]) is the difference between a calculated value and the real value. By dividing it by the real value you get the relative error. A bit relative error indicates bad calculation. | ||
+ | The largest relative error between two arrays as percentage: | ||
+ | <syntaxhighlight lang=python> | ||
+ | max(abs(calculated - real) / real ) * 100 | ||
</syntaxhighlight> | </syntaxhighlight> |
Revision as of 21:10, 24 December 2018
Mostly based on this paper that comes with its own modsim library.
Eyeopeners
- Store results in a list
for a in range(100):
funcAresults[a] = functionAcall(bla,bla)
funcBresults[a] = functionBcall(bla,bla)
- ModSimPy is using Series from Pandas to store results. This adds handy functions like.
- The state of the model is stored in a Pandas Series too.
- Put other interesting metrics in the state object too.
- If you use randomness in a function, get the mean values of several runs.
- Check the effect of different parameter values using Numpy linspace.
funcAresults = pd.Series([])
p1_array = np.linspace(0,1,12)
for p1 in p1_array:
for a in range(50):
funcAresults[a] = functionAcall(p1,p2)
print(funcAresults)
- The Approximation error (benaderingsfout) is the difference between a calculated value and the real value. By dividing it by the real value you get the relative error. A bit relative error indicates bad calculation.
The largest relative error between two arrays as percentage:
max(abs(calculated - real) / real ) * 100