The Access to Information Law obliges government agencies to release relevant data for citizens on their websites. The Chamber of Deputies, for example, displays expenses for each parliamentarian, but not everyone can easily access them.
It was with this in mind that Rayland Magalhães, a statistics student at the Federal University of Rio Grande do Norte (UFRN) created the available tool on this link. It allows you to view the refund requests of each federal deputy in a simple way.
To Tecnoblog, the student stated that the Chamber website fails to display data in isolation for the citizen. “One piece of information is not enough for him to draw a conclusion from what the deputy is doing”, he says.
For Rayland, public agencies should facilitate access to information. “The first step has already been taken: the data is open and anyone can access it. It would need to reduce the bureaucracy of this access to data so that anyone can access it easily and does not need intermediaries ”.
The page has graphics and highlights that indicate how much the deputy used his Parliamentary Quota during his term. The amount, which covers items such as airline tickets, fuel, accommodation and food, varies according to the state and reaches R $ 45,612.53 per month for parliamentarians in Roraima.
When typing a deputy’s name, the tool displays all of his expenses. They are detailed by year, type and company that performed the service. There is also a highlight that shows how the deputy’s expenditure compares with that of the other members of the House.
According to the student, requests for reimbursement demonstrate that the most common expenses among Brazilian deputies involve disclosure of parliamentary activity. At the same time, less common expenses are related to courses and specializations. “This shows a profile of the parliamentarians that we are electing,” he says.
The website, which has had requests for refunds made since 2009, adds a vast amount of information. “You cannot analyze this in an Excel spreadsheet, for example. There are more than 3 million rows and 50 columns. It is a lot, it is a very large volume of data and the software that ordinary people use, an Excel, is not yet able to analyze this easily ”, says the student.
The tool is part of a Rayland scientific initiation project at UFRN. To collect the data, he used the package that his advisor, professor Marcus Nunes, created with the programming language R. The package takes advantage of information collected by the module made in Python by Operation Serenade of Love, part of the Open Knowledge Brasil initiative.
Machine learning to monitor deputies
Rayland reports that the tool is still under development. The next step involves using machine learning to find possible fraud with parliamentary spending on points outside the curve.
“The idea is to create an outlier detection algorithm that is not supervised, that is, without the information that a refund request is fraud or is regular, we want to find out if that is a fraud or is in compliance”, explains.
According to the student, the tool would “analyze as many variables as possible and see which are different from the others, what stands out, what is aberrant”. Machine learning algorithms are still being analyzed and may be included in the tool in the future.
With information: UFRN.