CastaƱeda Molina, Eduar A.

Profile Picture

Filters

20231
Reset filters

Settings

Citations

Publication Search Results

Now showing 1 - 1 of 1
  • Thumbnail Image
    Publication
    An initialization algorithm using distribution-free method
    (2023-05-12) CastaƱeda Molina, Eduar A.; AlmodĆ³var Rivera, Israel A.; College of Arts and Sciences - Sciences; ColĆ³n Reyes, Omar; Lorenzo GonzĆ”lez, Edgardo; Department of Mathematics; Del Pilar Albaladejo, Joselyn
    Clustering is an unsupervised technique that partitions a dataset into homogeneous groups. The choice of initial values is a critical component in the performance of clustering algorithms. These values have a significant impact on the performance of these algorithms. In this study, we propose an initialization algorithm that combines the empirical likelihood approach with the normed residuals of the observations that have been chosen as initial values and their cumulative distribution function. Potential candidates for the initial values are the farthest from each other. Based on the empirical likelihood, these values will have a higher weight than those already considered. We prove that, if the initial values are obtained using our methodology, the expected objective function is reduced. Simulation experiments are carried out to study the proposed methodology. Our methodology is compared with popular initialization methods in terms of performance, that is, finding cluster solutions, as well as in terms of iterations. Our methodology is a top performer in finding homogeneous spherical groups, requiring a smaller number of iterations to converge than competing methods. Finally, the proposed methodology is applied to several real datasets.