Representaciones cromosómicas del algoritmo NSGA-II para el problema de árbol de expansión mínima multiobjetivo (MST-MO)
The purpose of this titration work was to identify the best alternative of chromosome representation that could improve the performance in multiobjective terms of a genetic algorithm known as NSGA-II, which was used at the same time to address the problem of the minimum spanning tree in its multiobj...
Saved in:
| Main Author: | |
|---|---|
| Format: | bachelorThesis |
| Language: | spa |
| Published: |
2019
|
| Subjects: | |
| Online Access: | http://repositorio.espam.edu.ec/handle/42000/1105 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Add Tag | Summary: | The purpose of this titration work was to identify the best alternative of chromosome representation that could improve the performance in multiobjective terms of a genetic algorithm known as NSGA-II, which was used at the same time to address the problem of the minimum spanning tree in its multiobjective version. This identification was achieved through the thorough evaluation of three different types of chromosome representation: Prüfer number, link and node biased encoding (LNB), and edge-set encoding. In conjunction with this, scientific research methods (bibliographic and analytical) were used to search, interpret and analyze the information and the results obtained in each execution. RStudio software was used for the coding and subsequent execution of the representations, which were applied to the NSGA-II together with their respective evolutionary operators. In the realization of the computational experiments, gene expression data were used, which correspond to four different databases available on the web: arabidopsis, cell cycle, sporulation and serum. Once these experiments were performed, the performance of the representations was evaluated using the hypervolume multiobjective performance metric, in order to identify the best Pareto boundaries of each representation. After the calculation of the hypervolume, statistical tests were applied to their data by means of the Friedman and Nemenyi tests. These hypervolume calculations and statistical tests allowed us to determine that the LNB representation was the best alternative for the improvement of NSGA-II. |
|---|