Yazar "Yibre, Abdulkerim Mohammed" seçeneğine göre listele

Listeleniyor 1 - 1 / 1
  • Küçük Resim
    Öğe
    Development of a novel candidate solution quality prediction approach to artificial algae algorithm
    (Selçuk Üniversitesi Fen Bilimleri Enstitüsü, 2020) Yibre, Abdulkerim Mohammed; Kıran, Mustafa Servet
    Metaheuristic optimization algorithms are capable of exploring solutions globally. This makes them attractive in the optimization of various real world problems. The maximum number of fitness evaluations (MaxFES) is one of the main factors shaping the success of an optimization algorithm, since it is expected to report acceptable result at reasonable time. But, it is not always possible that fitness evaluations are concluded with a successful fitness updates. In addition, the allowable MaxFES is restricted, and balancing exploration and exploitation is cumbersome in optimization algorithms. The algorithm should find best possible solution in an acceptable time. It is clear that more fitness computation requires more time. The performances of optimization algorithms are evaluated with predetermined MaxFES. The extent of execution for each fitness computation may differ in accordance with the problem. Because of that, obtaining best result with comparatively lesser fitness computation is challenging in optimization algorithms. To address these problems in this paper we proposed a novel approach named as AAANB that employs Gaussian-based Naïve Bayes probabilistic model to predict the quality of a candidate solution before an evaluation of its fitness value. The objective function is executed if its quality is predicted to generate good result. If not, the algorithm generates new candidate solution as usual. The main aim of this study is to improve the performance of AAA and apply it to solve continuous and real-world problems. The proposed method is applied in solving continuous optimization problem involving 26 standard benchmark functions and CEC'05 test suite. The comparative analysis showed that the proposed method surpassed Artificial Algae Algorithm (AAA) and other well-known optimization algorithms namely; Artificial Bee Colony (ABC), Particle Swarm Optimization (PSO), Bat Algorithm (BA), Flower Pollination Algorithm (FPA), Whale Optimization Algorithm (WOA) and Bee algorithm. The improved AAA has achieved best results with less function evaluation. AAANB is applied in training feed forwarded neural network (FFNN-AAANB) for predicting semen quality of males. The comparison of predictive performance of FFNN-AAANB revealed that, it can better identify normal and abnormal semen examples than MLP, KNN, SVM, NB, RF algorithms. In addition, AAANB is also used for constrained optimization of parameters of hybrid composite laminate with the purpose of obtaining minimum weight. The hybrid composite plate is exposed to twodirectional compressive forces. The composite design is subjected to a constraint called critical buckling load factor, which is a measure of how far the composite can withstand without buckling against the compressive forces.The numerical simulations have demonstrated that, AAANB can explore global solutions for continuous optimization problems, neural network weight optimization and hybrid composite laminate weight optimization with lesser function evaluations