The project carried out by a multidisciplinary team comprised of 4 AI doctoral students from UFPE and Economists from the Central Bank aimed to understand overindebtedness through the application of Machine Learning techniques. We utilized predictive models to forecast the risk of indebtedness, considering monitorable indicators over time. The interpretability of these models was crucial in understanding how variables affect overindebtedness. Additionally, we employed descriptive techniques to… Show more The project carried out by a multidisciplinary team comprised of 4 AI doctoral students from UFPE and Economists from the Central Bank aimed to understand overindebtedness through the application of Machine Learning techniques. We utilized predictive models to forecast the risk of indebtedness, considering monitorable indicators over time. The interpretability of these models was crucial in understanding how variables affect overindebtedness. Additionally, we employed descriptive techniques to identify new segments of overindebted individuals with distinct patterns of indebtedness. Our achievements included: a quantitative definition of overindebtedness, objective models to identify the phenomenon, the ability to predict individual risks, study of the phenomenon across different population segments, and an understanding of the characteristics that make individuals more prone to overindebtedness. Show less

Integrated into the data team, I was responsible for the churn modeling project (cancellation rate). I developed the ETL process to explore the data (EDA), built the models, and led their production. All work was carried out seamlessly within the DATABRICKS environment.

Integrated into the sugar cane field monitoring team, my main task involved developing anomaly detection models on satellite and drone images using deep learning models (ResNet). By analyzing these images, I contributed to providing the company with a comprehensive view of the extensive 300,000 acres of sugar cane fields. This monitoring allowed us to identify crop failures and evaluate various indicators of crop health.

Research and development of an data mining application using time series analysis and data visulization via Tableau for a power supply company.

Research based on optimizing time series forecasts using hybrid models and TRE algorithms for lag selection and hyperparameter tuning.

Pesquisa voltada para a aplicação Machine Learning e Análise de dados em problemas de engenharia civil.

PESQUISADOR VOLUNTARIO DE INICIAÇÃO CIENTÍFICA, PESQUISA: RADIER ESTAQUEADO NO NORDESTE BRASILEIRO.