Spatial and non-spatial clustering algorithms in the analysis of Brazilian educational data

Abstract

Education is one of the pillars of human societies, such that achieving better indicators in this area is a common goal for different federate entities. In this context, identifying patterns on the results of such indicators, evaluated for different entities, as well as grouping them based on their similarities, can lead to a better understanding of the educational scenario of a population. This knowledge, moreover, might subsidize the formulation of public policies and allow the decision-making by the responsible managers. In the present work, we present an illustrative example of the application of spatial and non-spatial clustering algorithms in the analysis of data from six important indicators of basic education (middle and high school) evaluated for the municipalities of the state of Paraná, Brazil. Clusters provided by each method were evaluated according to their spatial distributions and educational features. The different clustering algorithms produced clusters with different levels of spatial contiguity and homogeneity regarding the educational indicators, reflecting the importance of choosing the appropriate clustering technique based on the research objectives.

KEYWORDS:

• ClustGeo algorithm
• cluster inertia
• hierarchical clustering
• minimum spanning tree
• non hierarchical clustering
• skater algorithm