Тип публикации: доклад, тезисы доклада, статья из сборника материалов конференций
Конференция: Springer Science and Business Media Deutschland GmbH; 6 July 2020 through 10 July 2020; 6 July 2020 through 10 July 2020
Год издания: 2020
Идентификатор DOI: 10.1007/978-3-030-58657-7_28
Ключевые слова: clustering, genetic algorithm, greedy agglomerative procedure, k-means
Аннотация: Progress in the development of automatic grouping (clustering) methods, based on solving the p-median and similar problems, is mainly aimed at increasing the computational efficiency of the algorithms, their applicability to larger problems, accuracy, and stability of their results. The researchers’ efforts are focused on the develПоказать полностьюopment of compromise heuristic algorithms that provide a fairly quick solution with minimal error. The Genetic Algorithms (GAs) with greedy agglomerative crossover procedure and other special GAs for the considered problems demonstrate the best values of the objective function (sum of squared distances) for many practically important problems. Usually, such algorithms do not use any mutation operator, which is common for other GAs. We propose new GAs for the k-means problem, which use the same procedures as both the crossover and mutation operators. We compared a simple GA for the k-means problem with one-point crossover and its modifications with the uniform random mutation and our new crossover-like mutation. In addition, we compared the GAs with greedy heuristic crossover procedures to their modifications which include the crossover-like mutation. The comparison results show that the idea of our new mutation operator is able to improve significantly the results of the simplest GA as well as the genetic algorithms with greedy agglomerative crossover operator. © 2020, Springer Nature Switzerland AG.
Журнал: Communications in Computer and Information Science
Выпуск журнала: Vol. 1275 CCIS
Номера страниц: 350-362
ISSN журнала: 18650929