Metaheuristic Enhancement with Identified Elite Genes by Machine Learning

Document Type

Conference Proceeding

Publication Date



The traveling salesman problem (TSP) is a classic NP-hard problem in combinatorial optimization. Due to its difficulty, heuristic approaches such as hill climbing (HC), variable neighborhood search (VNS), and genetic algorithm (GA) have been applied to solve it as they intelligently explore complex objective space. However, few studies have focused on analyzing the objective space using machine learning to identify elite genes that help designing better optimization approaches. For that, this study aims at extracting knowledge from the objective space using a decision tree model. According to its decision-making basis, a simulated boundary is reproduced to retain elite genes from which any heuristic algorithms can benefit. Those elite genes are then integrated into a traditional VNS, unleashing a remarkable enhanced VNS named genetically modified VNS (GM-VNS). Results show that the performance of GM-VNS surpasses conventional VNS in terms of solutions’ quality on various real-world TSP instances.

Publication Title

Communications in Computer and Information Science

First Page Number


Last Page Number




This document is currently not available here.