Geoffrey De Smet created JBRULES-3563:
-----------------------------------------
Summary: Planner: Benchmark report should also show percentage versus worst
dataset (and an average of that percentage)
Key: JBRULES-3563
URL:
https://issues.jboss.org/browse/JBRULES-3563
Project: Drools
Issue Type: Feature Request
Security Level: Public (Everyone can see)
Components: drools-planner
Reporter: Geoffrey De Smet
Assignee: Geoffrey De Smet
Managers like to know much better a result it in relative terms.
This often boils down to comparing the results of the construction heuristic with the
results of 1 or more full-fledged configuration (construction heuristic + metaheuristic).
So for each dataset for each solver configuration, we can calculate a percentage: ((-
Worst solver's score) - (- this solver's score)) / (- Worst solver's score).
For the Worst's solver's score this will be 0%.
If the worst solvers score is 0, then it should be NaN or empty.
Then, for each solver configuration, we can calculate an "average percentage".
Note that is is calculated based on the percentages, not the average of the scores! If any
datasets have NaN or empty, that result should be discarded (which means the entire column
gets discarded).
Classes to look into:
- SingleBenchmark should have the betterThanWorstScorePercentage. TODO: find a better name
then that :)
- SolverBenchmark should have the averageBetterThanWorstScorePercentage TODO: find a
better name then that :)
- PlannerStatistic should output a graph of it. Copy paste from
writeBestScoreSummaryChart. (add it after the WinningScoreDifference code)
- index.html.ftl should write the numbers in the report (add it after
WinningScoreDifference)
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.jboss.org/secure/ContactAdministrators!default.jspa
For more information on JIRA, see:
http://www.atlassian.com/software/jira