Proposed methodology for the creation of a classification label: a school performance case study

Abstract

Quality in Education is an issue that has been discussed in schools and among their managers, in the media, and in the literature. However, a deeper review of the literature has failed to present techniques dealing with database information techniques capable of obtaining classifications for school performance; nor is there a consensus regarding the definition of “educational quality”. To address the situation, in this paper, we propose a methodology that fits the KDD (Knowledge Discovery in Databases) process to classify teaching in schools. This is done by comparing the grades of the “Prova Brasil”, which is part of the Development Index of Basic Education (IDEB) in Brazil. To illustrate the methodology, it was applied to 17 public elementary schools in the municipality of Araucária, located in the metropolitan region of Curitiba, Parana state. The grades achieved by all students of the initial years (1^st to 5^th year of fundamental teaching) and final years (6^th to 9^th years of fundamental teaching) were considered. In the Data Mining phase, the main phase of the KDD process, three techniques were used comparatively: Artificial Neural Networks, Support Vector Machines, and Genetic Algorithms. Those techniques presented acceptable results in classifying each school represented by a “Performance Classification Label”. Based on this label, the educational managers can have a greater input for procedures to be adopted in each school, and thus set more accurate targets.

Keywords:
School performance; KDD process; Pattern recognition; Real case study