Тип публикации: монография
Год издания: 2024
Аннотация: A problem of increasing the accuracy and stability of the automatic grouping (clus-tering) algorithms is considered based on a new approach to developing clustering al-gorithms based on parametric optimization models for k-means, k-medoid, clear clus-tering problems based on separation of a mixture of probability distributions (witПоказать полностьюh ap-plication of the classification EM-algorithm). The study proposes new search algo-rithms with alternating randomized neighborhoods and parallel modifications of algo-rithms with a greedy agglomerative heuristic procedure for large automatic grouping problems, adapted to the CUDA architecture. Moreover, the study presents a procedure for composing optimal ensembles of automatic grouping algorithms with a combined application of the genetic algorithm of the greedy heuristic method and a consistent matrix of binary partitions for practical problems. Algorithms and procedure for com-posing optimal ensembles are implemented to solve the problem of dividing prefabri-cated lots of industrial products into homogeneous lots based on the results of non-destructive tests. The work is intended for scientists, specialists, undergraduate and postgraduate stu-dents involved in the development of cluster analysis algorithms, as well as in improv-ing the quality of industrial products.
Место издания: Moscow