Interpretable random forest model for identification of edge 3-uncolorable cubic graphs

Dudáš, Adam, and Modrovičová, Bianka. "Interpretable random forest model for identification of edge 3-uncolorable cubic graphs." Kybernetika 59.6 (2023): 807-826. <http://eudml.org/doc/299198>.

@article{Dudáš2023,
abstract = {Random forest is an ensemble method of machine learning that reaches a high level of accuracy in decision-making but is difficult to understand from the point of view of interpreting local or global decisions. In the article, we use this method as a means to analyze the edge 3-colorability of cubic graphs and to find the properties of the graphs that affect it most strongly. The main contributions of the presented research are four original datasets suitable for machine learning methods, a random forest model that achieves $97.35\%$ accuracy in distinguishing edge 3-colorable and edge 3-uncolorable cubic graphs, and the identification of crucial features of graph samples from the point of view of its edge colorability using Shapley values.},
author = {Dudáš, Adam, Modrovičová, Bianka},
journal = {Kybernetika},
keywords = {random forest; proper edge coloring; interpretable machine learning; snark},
language = {eng},
number = {6},
pages = {807-826},
publisher = {Institute of Information Theory and Automation AS CR},
title = {Interpretable random forest model for identification of edge 3-uncolorable cubic graphs},
url = {http://eudml.org/doc/299198},
volume = {59},
year = {2023},
}

TY - JOUR
AU - Dudáš, Adam
AU - Modrovičová, Bianka
TI - Interpretable random forest model for identification of edge 3-uncolorable cubic graphs
JO - Kybernetika
PY - 2023
PB - Institute of Information Theory and Automation AS CR
VL - 59
IS - 6
SP - 807
EP - 826
AB - Random forest is an ensemble method of machine learning that reaches a high level of accuracy in decision-making but is difficult to understand from the point of view of interpreting local or global decisions. In the article, we use this method as a means to analyze the edge 3-colorability of cubic graphs and to find the properties of the graphs that affect it most strongly. The main contributions of the presented research are four original datasets suitable for machine learning methods, a random forest model that achieves $97.35\%$ accuracy in distinguishing edge 3-colorable and edge 3-uncolorable cubic graphs, and the identification of crucial features of graph samples from the point of view of its edge colorability using Shapley values.
LA - eng
KW - random forest; proper edge coloring; interpretable machine learning; snark
UR - http://eudml.org/doc/299198
ER -