A Novel Genetic Algorithm Selection Method and Implementation in IoT Domain

Abstract

In Internet of Things (IoT) domain, there are many NP-hard problems that are required to be solved efficiently by some meta-heuristic algorithms such as genetic algorithm. There are many genetic algorithm selection methods which aim to find the optimal chromosome by populating and varying them by some mutation and cross-over methodologies. Most of these algorithms process all of the genes of chromosomes into fitness function to decide whether to pass them to new generation or to apply some genetic operations or to abandon them. If chromosomes are ultimately long or fitness functions are difficult to compute, it will have a great overhead and will consume so much time. In this paper, a novel genetic algorithm selection method is proposed and promises some intelligence to keep these processes short and efficient by taking some remarkable risks. It is based on analogy of a football league with multiple groups and it uses the idea of partially processing the chromosome genes after adopting a decision mechanism based on Bayesian game theory. We implement the proposed selection algorithm to solve a complex problem in IoT domain and illustrate its performance compared to other generic selection methods.