Characterization method for more than three dimensions based on dependence and correlation using hybrid multiple correspondence analysis

Abstract

Characteristics of a district refer to specific attributes that describe its condition and quality. Identifying these characteristics provides more precise insights into districts that require more attention in specific situations. These characteristics are determined based on the interdependence of categories. Dependencies among multiple qualitative and quantitative variables can be analyzed using multiple correspondence analysis (MCA) hybrid with cosine correlation. MCA is particularly suited for analyzing contingency tables with more than two qualitative variables. This study aims to identify the characteristics of objects based on the categories of qualitative variables and the characteristics of objects based on quantitative variables. Several qualitative variables with many categories may result in less representative information, potentially failing to reach 80% cumulative variance in two dimensions. Therefore, the novelty of this study lies in identifying characteristics using Euclidean distance with a variance percentage of 100% in more than three dimensions and used two types of variables. The data used in this study are eight qualitative variables with 49 categories and three quantitative variables from the Supporting Area Survey of Bandung Regency in 2023. The analysis results indicate that the cumulative variance in two dimensions is only 16.9%. Consequently, calculations were performed using an Euclidean distance matrix with 43 dimensions to achieve 100% cumulative variance. The results of the analysis revealed that 28 district groups were identified.