Analyzing hybrid implementations of optimization algorithms

Abstract

<p>Optimization problems can be found in almost every aspect of the practical world. They can range from something as complex as a wireless communication system or the stock exchange to something as mundane as scheduling a series of meetings inside a company. Almost every task in today’s world can be reconceived as a type of optimization problem. To solve these problems, entire fields in computer intelligence have been developed and continue to be the focus of extensive research today. Various algorithms have been proposed that effectively find solutions and do so in the shortest amount of time possible. However, the number of problems far exceeds the number of algorithms that have been developed to combat them. As such, each optimization algorithm is expected to produce optimal or at least near-optimal solutions in many optimization situations. Whilst this is certainly possible, various theorems state this to be unlikely. Thus, there is always a need to create new optimization algorithms. in recent years, a possible solution has been proposed; to combine the favorable characteristics of two optimization algorithms into a hybrid implementation.</p> <p>This MRP investigates the performance of various hybrid implementations of optimization algorithms in various optimization scenarios. Utilizing thirteen benchmark functions, it compares the performance of seven algorithms; Genetic Algorithm (GA), Particle Swarm (PSO), Grey Wolf (GWO), Whale Optimization (WOA), Binary Bat Algorithm (BAT), a Binary Bat/Particle Swarm hybrid (BATPSO), and a Grey Wolf/Whale Optimization (GWOWOA) hybrid. This comparison is made over several case scenarios which tweak the general parameters found in these metaheuristic algorithms. To further quantify the results, statistical hypothesis testing techniques are utilized to differentiate between the performance of each algorithm. These tests determine the comparative gap in the performance samples gathered from each metaheuristic. Finally, it compares the performance of Genetic algorithm (GA) and Particle Swarm PSO) on a Fuzzy Inference System (FIS) tree application to gauge how optimization algorithms play a significant role in many real-world systems.</p> <p>In its conclusion, it presents a summary of the findings made and recommends future courses of</p> <p>action.</p>