Cluster Analysis Using Ensemble ROCK Method in District/City Grouping in South Sulawesi Province based on People's Welfare Indicators

Taufiq  Hidayat; Ruliana Ruliana; Zulkifli Rais; Miguel Botto-Tobar

doi:10.35877/mathscience1761

Authors

Taufiq Hidayat Department of Statistics, Universitas Negeri Makassar, Makassar, 90223, Indonesia
Ruliana Department of Statistics, Universitas Negeri Makassar, Makassar, 90223, Indonesia
Zulkifli Rais Department of Statistics, Universitas Negeri Makassar, Makassar, 90223, Indonesia
Miguel Botto-Tobar (1) Eindhoven University of Technology, 5600 MB Eindhoven, The Netherlands; (2) Research Group in Artificial Intelligence and Information Technology, University of Guayaquil, 090510, Guayaquil, Ecuador

DOI:

https://doi.org/10.35877/mathscience1761

Keywords:

Data Mining, Cluster Analysis, ROCK, Agglomerative Hierarchy, Cluster Ensemble ROCK

Abstract

Cluster analysis is a data mining technique used to group data based on the similarity of attributes of object data. One of the problems that are often encountered in cluster analysis is data with a mixed categorical and numerical scale. The clustering stage for mixed data using the ensemble ROCK (Robust Clustering using links) method is carried out by combining clustering outputs from categorical and numeric scale data. The method used for categorical data is the ROCK method and the method used for numerical data is the Hierarchical Agglomerative method. The best clustering method is determined based on the criteria for the ratio between the standard deviations within the group (SW) and the smallest standard deviation between groups (SB). Based on 24 observation objects in the regencies and cities of the Province of South Sulawesi, the ROCK ensemble method with a value of 0.1 produces three clusters with a ratio value of 2,27 x10-16 based on the combination of the output results of the ROCK method and the Hierarchical Agglomerative method