Performance improvements of social spider algorithm in continuous and discrete optimization problems

Evolutionary calculation based on natural phenomena can be divided into two important groups. These are evolutionary algorithms and intelligent swarm-based algorithms. Evolutionary algorithms are algorithms inspired by nature. It can achieve successful results in evolutionary optimization problems and real world problems. Such algorithms include genetic algorithm (GA), genetic programming (GP), evolutionary strategies (ES), and differential evolution (DE). Intelligent swarm based algorithms have attracted a lot of attention in recent years. Such algorithms include the Particle Swarm Optimization algorithm (PSO), the Artificial Bee Colony algorithm (ABC), the Ant Colony Optimization algorithm (ACO), and the Bat Algorithm (BA). The term swarm refers to the community of individuals who are in contact with each other. Intelligent swarm-based algorithms imitate the behavior of bird, ant, bee, bacteria, butterfly, spider etc. Intelligent swarm-based algorithms can successfully solve many different types of problems such as continuous, discrete and binary optimization problems. They are very successful in solving the problem types of NP-hard which are very difficult to solve. The newly developed Social Spider Algorithm (SSA) in recent years is a swarm-based algorithm. It is a swarm based algorithm created by imitating spider species living together in nature. Within the scope of this thesis, the success of SSA on continuous, discrete and binary optimization problems is examined and developed by adding new methods to increase its success. Social Spider Algorithm (SSA) is a newly developed herd-based algorithm in recent years. It was created by imitating spider species that live together in nature. Within the scope of this thesis, the success of SSA on continuous, discrete and binary optimization problems has been examined and SSA has been developed by adding new methods to increase its success. The SSA was tested in small size (10, 20, and 30) continuous optimization problems in the original algorithm for which SSA was proposed first. In this thesis, the original version of SSA was developed by adding spider explosion and explorer spider memory features and Social Spider Algorithm (ISSA) was proposed. In addition, the ability to search locally and globally has increased. The success of ISSA in low, middle, and high dimensional (10, 20, 30, 100, 500, and 1000) continuous optimization problems has also been tested. In a second study performed for continuous optimization tasks, the success of ISSA on MPEF was examined by adding four features to the original SSA in the form of crossover, mutation, Gbest convergence and silent spider. MPEF is a scalable, simplified Molecular Potential Energy Function. In the third study, the proposed SSA focusing on continuous optimization problems was transformed into a discrete form in order to solve the discrete optimization problems known as the problem types that take independent discrete values, and the Discrete Social Spider Algorithm (DSSA) was proposed. The Traveling Salesman Problem (TSP), which is a discrete optimization problem frequently preferred in the literature, has been solved with DSSA. In DSSA, exploration and exploitation capabilities are enhanced by adding skilled spider and novice spider features. In this thesis, binary optimization problems, a subset of discrete optimization, are also solved. With the Binary Social Spider Algorithm (BinSSA) proposed in this study, four different binary optimization problems (feature selection problem, uncapacitated facility location problem, wind turbines placement problem and continuous tasks) are solved. In this study, S-shaped, V-shaped and mode-based transfer functions are used while converting the continuous search space to a binary search space. In order to support the success of transfer functions in binary optimization, logic gates (xor logic gate), similarity measurement techniques (Jaccard and Dice), and the crossover operator, which are frequently used in recent years, have been preferred to obtain new candidate solutions. Thanks to these methods, BinSSA can discover new points in the binary search space or find new points around local points. Thus, the success of BinSSA has been improved.