Decision Trees in Proper Edge k-coloring of Cubic Graphs

This work examines the possibilities of increasing the efficiency of the computation of proper edge k-coloring of cubic graph with the use of machine learning methods. State-of-the-art approaches related to this problem work with time complexity of circa O(2^0.427|V(G)|), where| V(G)| is number of vertices of given graph G. The main focus of the paper is use of machine learning model of decision trees for the problem of identification of properly edge 3-uncolorable graphs called snarks - well known instance of NP-complete problem. Presented work consists of creation of graph property datasets fitting for the specified machine learning task, building of decision tree models based on the created datasets and identification of properties which are significant in the context of graph edge coloring and are measurable in lower time complexity than the edge coloring itself.