ANALISIS DISKRIMINAN UNTUK VALIDASI CLUSTER PADA STUDI KASUS PENGELOMPOKAN KECAMATAN DI KABUPATEN JEMBER BERDASARKAN STATUS KEMISKINAN

Discriminant Analysis for Cluster Validation in the Case Study of District Grouping in Jember Regency Based on Poverty Status

Abstract

. Cluster validation is a procedure to evaluate the results of cluster analysis quantitatively and objectively on a data. The validation process is very important to get the results of a good and appropriate grouping. In the validation process, the author uses internal validation, stability, and discriminant analysis test. This study aims to obtain validation results from the hierarchy and kmeans method. This data grouping uses “iris” simulation data, which results from the grouping method used can be applied to the original data to see which validation method is used for all data and produce an optimal grouping. The result of the study show that in the “iris” data, a single linkage link is an appropriate grouping method because the result of the grouping are optimal for all validations and classification of group members whose groups are significant. In Sub-district poverty data in Jember district with a single linkage link optimal grouping was obtained and complete linkage links were also used as a method that resulted in optimal grouping for all validation. Cluster validation using discriminant analysis test is appropriate for various types of data in general and shows that single linkage methods are better than other methods for grouping and validation methods for “iris” data and Sub-district data in Jember district based on variables of poverty status.

Keywords: Cluster Analysis, Diskriminant Analysis, Multivariate Analysis, Validation Cluster