Hybridizing Meta-RaPS with Machine Learning Algorithms

Abstract

Merging a metaheuristic with machine learning algorithms is typically done to improve the machine learning algorithms. This work, however, takes the reverse approach and aims at utilizing machine learning algorithms to improve metaheuristics. The objective of this research is to demonstrate an effective approach to hybridize metaheuristics with machine learning. The metaheuristic of choice is Metaheuristic for Randomized Priority Search (Meta-RaPS) and the machine learning algorithms are Decision Trees (supervised learning) and Association Rules (unsupervised learning). Demonstrating the performance of the algorithms is done by solving the Vehicle Routing Problem (VRP). This paper starts by describing the Vehicle Routing Problem and then subsequent sections discuss the algorithms used and the computational experiments executed.