Pattern Identification Using Fuzzy Cluster Analysis and Latent Class Analysis: A Case Study in Perú

Jorge Chue  Gallardo; César Higinio  Menacho Chiock; Jesús Walter Salinas  Flores; Iván Dennys Soto  Rodríguez; Raphael Félix Valencia  Chacón; Rino Nicanor Sotomayor  Ruiz; Fernando Rene  Rosas Villena; Frida Rosa Coaquira  Nina

doi:10.36941/ajis-2024-0111

Authors

Jorge Chue Gallardo Academic Department of Statistics and Informatics, Universidad Nacional Agraria La Molina, Av. la Molina s/n, La Molina, Perú
César Higinio Menacho Chiock Academic Department of Statistics and Informatics, Universidad Nacional Agraria La Molina, Av. la Molina s/n, La Molina, Perú
Jesús Walter Salinas Flores Academic Department of Statistics and Informatics, Universidad Nacional Agraria La Molina, Av. la Molina s/n, La Molina, Perú
Iván Dennys Soto Rodríguez Academic Department of Statistics and Informatics, Universidad Nacional Agraria La Molina, Av. la Molina s/n, La Molina, Perú
Raphael Félix Valencia Chacón Academic Department of Statistics and Informatics, Universidad Nacional Agraria La Molina, Av. la Molina s/n, La Molina, Perú
Rino Nicanor Sotomayor Ruiz Academic Department of Statistics and Informatics, Universidad Nacional Agraria La Molina, Av. la Molina s/n, La Molina, Perú
Fernando Rene Rosas Villena Academic Department of Statistics and Informatics, Universidad Nacional Agraria La Molina, Av. la Molina s/n, La Molina, Perú
Frida Rosa Coaquira Nina Academic Department of Statistics and Informatics, Universidad Nacional Agraria La Molina, Av. la Molina s/n, La Molina, Perú

DOI:

https://doi.org/10.36941/ajis-2024-0111

Keywords:

Pattern identification, fuzzy cluster analysis, latent class analysis, Dunn's fuzziness coefficient

Abstract

The Demographic and Family Health Survey (ENDES) conducted by the National Institute of Statistics and Informatics (INEI) in Peru provides data on fertility and health. The ENDES 2020 report, based on 35,847 surveyed households, undergoes descriptive statistical analysis with the aim of identifying patterns to enhance social conditions. Techniques such as Fuzzy C-Means and Latent Classes, previously applied in various contexts, are employed. Correlation analysis using the R polycor package highlights significant relationships, leading to the exclusion of certain numeric variables in fuzzy clustering due to strong correlations. Random sampling is applied to address the data volume. Three clusters are determined through kmeans clustering, silhouette, Elbow, and Clara methods, assessing their fuzziness with the Dunn's Fuzziness Coefficient. Pattern identification reveals significant differences in family relationships, gender, education, and health insurance among the clusters. The widespread lack of health insurance, particularly ESSALUD/IPSS, stands out as a common issue. Fuzzy clustering and latent class analysis techniques provide groupings with variations in sizes and compositions.

Received: 20 March 2024 / Accepted: 28 June 2024 / Published: 02 July 2024