Performance metrics in multi-objective optimization
Nery Riquelme$^{1}$, Christian Von Lücken$^{1}$, Benjamín Barán$^{2}$
$^{1}$Universidad Nacional de Asunción. San Lorenzo Paraguay,
$^{2}$Universidad Nacional del Este. Paraguay
email: neryriquelme90@gmail.com, clucken@pol.una.py, bbaran@cba.com.py

Schedule:Mon 19th@17:45, Room: A

In the last decades, a large number of metrics has been proposed to compare the performance of different evolutionary approaches in multi-objective optimization. This situation leads to difficulties when comparisons among the output of different algorithms are needed and appropriate metrics must be selected to perform those comparisons. Hence, no complete agreement on what metrics should be used exists. This paper presents a review and analysis of 54 multi-objective-optimization metrics in the specialized literature, discussing the usage, tendency and advantages/disadvantages of the most cited ones in order to give researchers enough information when choosing metrics is necessary. The review process performed in this work indicates that the hypervolume is the most used metric, followed by the generational distance, the epsilon indicator and the inverted generational distance.

BibTex

@InProceedings{CLEI-2015:144491,
	author 		= {Nery Riquelme and Christian Von Lücken and Benjamín Barán},
	title 		= {Performance metrics in multi-objective optimization},
	booktitle 	= {2015 XLI Latin American Computing Conference (CLEI)},
	pages 		= {286--296},
	year 		= {2015},
	editor 		= {Hector Cancela and Alex Cuadros-Vargas and Ernesto Cuadros-Vargas},
	address 	= {Arequipa-Peru},
	month 		= {October},
	organization 	= {CLEI},
	publisher 	= {CLEI},
	url 		= {http://clei.org/clei2015/144491},
	isbn 		= {978-1-4673-9143-6},
	}