Relationship of federal funding to IDEB results in a state in Brazil: an approach based on Educational Data Mining

摘要

To ensure transparency and provide ways for analysis and auditing of various public activities and expenditures, the Complementary Law 131, known as the "Transparency Law," requires the Union, States, and Cities to disclose their expenses in real-time through Internet. However, despite the apparent progress in opening data in Brazil, applications do not follow the various recommended models. Since this data is mostly unconnected, a clear view of the underlying contexts of a query becomes impossible. Thus, there is no clear relationship between recorded expenditure data for education, quality indicators, and/or descriptive census data. The following research sought to create relations between data from the Brazilian Federal Government Open Data Portal and the primary education quality indicator, IDEB, in the cities of the state. The methodology used was Data Mining, a research field resulting from the intersection between the areas of Computer Science and statistics, which seeks to discover non-trivial information hidden in large data. Using Correlation and Linear Regression in the collected and formatted data, it was not found linear relationship, indicating that only cities funding data is not sufficient to obtain a causal relationship with the IDEB average for municipalities. The statistical result found, although it does not provide a direct correlation between the variables investigated, represents a relevant scientific finding in the non-causal relationship itself. The more relevant contribution of this research is the process of selection, preprocessing, and transformation of the data scattered in different information sources and different formats of the analyzed data, as well as the originality of the analyzed causality.