Cluster analysis applied to identify the families with similar housing characteristics among those enrolled in the Unified Registry in the municipality of Brumadinho - Minas Gerais

The precarious settlements, dwellings with inadequate infrastructure where low-income families live, are a challenge to the right to housing in Brazil, affecting about 5.8 million households. The study analyzed the housing characteristics of families enrolled in the Unified Registry (Cadastro Único) in the municipality of Brumadinho - MG, seeking to identify similar groups among them. The housing conditions of each family were summarized in the form of the Minas Gerais Housing Vulnerability Index (IMV-H) composed of two indicators: the Housing Inadequacy, which measures the quality of the home's infrastructure, and the Housing Burden, which measures the weight of housing expenses in the family's monthly income. Conglomerate analysis techniques were applied to form groups of similar families by implementing the K-means and K-medoids methods. The comparison between the different methods indicated the Housing burden as the determining variable for the differentiation of families, and the K-means algorithm was the one that generated the best clustering results. Thus, the families were divided into two groups: group 1, formed by families who spent 55% or more of their income on housing expenses, and group 2, formed by families who spent up to 55% of their income on these expenses. The Silhouette Coefficient found for the clusters was 0.71, indicating a solid clustering structure. The families and their clusters were geolocalized from spatial data of the public equipment associated with them in the Unified Registry.