Yazar "Akdağlı, Ali" seçeneğine göre listele

Listeleniyor 1 - 2 / 2
  • Küçük Resim
    Öğe
    Designing of the Artificial Neural Network Model Trained by Using the Different Learning Algorithms to Classify the Electrocardiographic Signals
    (European Journal of Science and Technology, 2022) Çelik, İbrahim; Üstün, Deniz; Akdağlı, Ali
    An artificial neural network model trained by using various learning algorithms is designed to classify the electrocardiographic signals in this study. The model of artificial neural network is constructed on the structure consisting of a multilayered perceptron based on the feed forward back propagation. A data pool is built by using a dataset consists of 66 electrocardiographic data’s taken from the MIT BIH arrhythmia database to perform the training and testing processes of artificial neural network model. The training process of artificial neural network model is performed with 46 electrocardiographic data and then the accuracy of the model is tested via 20 electrocardiographic data. The artificial neural network is trained by 3 different learning algorithms to achieve a robust model. The performance of the learning algorithms used for training the model of the artificial neural network is evaluated according to percentage error. It illustrates that the artificial neural network model trained by Levenberg–Marquardt learning algorithm obtains the better classification result than other learning algorithms. The proposed artificial neural network model can be successfully used to classify the electrocardiographic signals.
  • Küçük Resim
    Öğe
    Modified artificial bee colony algorithm with differential evolution to enhance precision and convergence performance
    (Elsevier, 2022) Üstün, Deniz; Toktaş, Abdurrahim; Erkan, Uğur; Akdağlı, Ali
    Artificial bee colony (ABC) and differential evolution (DE) are the most powerful and operative meta-heuristic algorithms inspired by the nature. Although both algorithms are successful, their successes vary from phase to phase, i.e. while ABC is better in the exploration ability, DE is well in the exploitation capability. Because the diversity of mutation and exponential crossover operators is prominently better than that of onlooker bee; in this study, the exploitation ability of ABC is enhanced by replacing the onlooker bee operator with those of mutation and the crossover phases of DE in order to increase the accuracy and speed up the convergence. We hereby introduce a novel modified algorithm denoted “modified ABC by DE” (mABC). The precision performance of mABC is verified through 20 classical benchmark functions and CEC 2014 test suit by a comprehensive comparison with recent ABC variants and hybrids for 30 and 50 dimensions. The results are interpreted using various statistical evaluations such as Wilcoxon, Friedman, and Nemenyi tests. Moreover, mABC is comparatively examined over convergence plots. In concise, the mean ranks of mABC are 1.4 and 2.3 for classical benchmark functions and CEC 2014, respectively. mABC outperforms the other variants averagely for 14 of 20 classical benchmark functions and 24 of 30 CEC 2014 functions. The results manifest that the proposed mABC is a robust and reliable algorithm as well as better than the existing ABC variants and hybrids with regard to high optimization performance like precision and convergence