OPTIMIZATION OF K-MEANS WITH SIMULATED ANNEALING HEURISTICS
Ivan O. Cruz-Ruiz. Universidad Autónoma Metropolitana Unidad Iztapalapa, Av. San Rafael Atlixco 186, 09340, Ciudad de México, México. Email: iocr@xanum.uam.mx
Pedro Lara-Velázquez. Universidad Nacional Autónoma de México, C.U., Coyoacán, 04510, Ciudad de México, México. plara@xanum.uam.mx
Eric A. Rincón-García. Universidad Nacional Autónoma de México, C.U., Coyoacán, 04510, Ciudad de México, México. rincon@xanum.uam.mx
Miguel A. Gutiérrez-Andrade. Universidad Nacional Autónoma de México, C.U., Coyoacán, 04510, Ciudad de México, México. gamma@xanum.uam.mx
Sergio G. de-los-Cobos-Silva. Universidad Nacional Autónoma de México, C.U., Coyoacán, 04510, Ciudad de México, México. cobos@xanum.uam.mx
Roman A. Mora-Gutiérrez. Universidad Autónoma Metropolitana Unidad Azcapotzalco, Av. San Pablo Xalpa 180, 02200, Ciudad de México, México. mgra@correo.azc.uam.mx
Carlos A. Hernández-Nava. Universidad Autónoma Metropolitana Unidad Iztapalapa, Av. San Rafael Atlixco 186, 09340, Ciudad de México, México. Email: cahn@xanum.uam.mx
- Fuzzy Economic Review: Volume 26, Number 2, 2021, special issue
- DOI: 10.25102/fer.2021.02.03
Abstract
The classification algorithm known as k-means is widely used in different branches of artificial intelligence. Its main objective is to find the k centroids that can repre-sent clusters of information. While this algorithm is reliable, it is not always possi-ble to find the optimal centroids that represent the data set. In this work, different metaheuristic techniques were applied in which the problem of finding the initial centroids was solved, and a novel hybrid algorithm is presented that finds better centroids than those found by the standard k-means algorithm. The new approach improves the classification of the information contained in the tested databases.