Students present final papers on Machine Learning

Students from the Machine Learning course, supervised by Professor Rodrigo Ribeiro, presented their final projects on Thursday (13). The proposals apply artificial intelligence algorithms and data analysis to different practical contexts. The subject, which is taken by undergraduates from the Data Science and Computer Science emphases, is part of the second year of the bachelor’s degree in Mathematics, Technology and Innovation.

The interdisciplinary work consisted of analyzing real data to develop mathematical models for classification and regression, with the aim of achieving good performance metrics. “We have a strong mathematical bias: we study what’s behind the algorithms, but we also get hands-on. The students collected data, programmed and implemented algorithms, making simulations with real information,” explained the professor.

Among the topics chosen by the undergraduates were data on public health, salaries for different professions and income distribution. Student Bianca Zavadisk’s group analyzed medical information to identify predictors capable of estimating the likelihood of a patient having diabetes. “It’s very interesting, especially in the medical field, to optimize processes. In the public health system, there is a lack of professionals and time, and sometimes a simple test or demographic data can quickly indicate a result. Mathematical models can help in various social areas,” he explained.

O dataset – a set of organized data for analysis and modelling – used by the group required a careful stage of cleaning and selecting information. “First, we separated and filtered out the most useful data for training the model. Then we applied theories, compared results and adjusted parameters to improve performance,” said the student.

Student Igor Roberto Alves’ group chose to work with information on heart disease. “We wanted to work with real data that has an impact on everyday life.” To do this, the students analyzed laboratory tests, family and clinical history and demographic data. “The biggest challenge was to minimize errors in patient classification while maintaining good metrics. An incorrect diagnosis can generate additional costs and overburden the public health system. An accurate model helps to avoid this problem,” explained Igor.

The projects involved all the stages of professional work, from collecting data to implementing algorithms. For teacher Rodrigo Ribeiro, the result demonstrates the students’ technical and social maturity. “The students managed to combine mathematical rigor and social awareness. When analyzing a set of data, it’s essential to ask relevant questions – and this often goes beyond theory. At IMPA Tech, we want to train mathematicians capable of solving concrete problems in society,” he concluded.

Read more: ‘IMPA Tech is a unique opportunity’, says student Mariana Yoshioka
See also: Undergraduates present scientific posters in English