Author: Rafaelli Coutinho

Data Analysis is a multidisciplinary field focused on interpreting large volumes of information to support decision-making, strategy development, and innovation. Statistical and machine learning techniques are employed to identify patterns and forecast future events, encompassing structured, semi-structured, and unstructured data.

For structured data, the main challenges involve analyzing time series and spatiotemporal data, including prediction, pattern discovery, and adaptation to data drift. Methods such as filtering and decomposition are used to build robust models for forecasting. The detection of events in time series, such as anomalies and regime changes, is relevant for both retrospective and real-time analysis.

When dealing with semi-structured and unstructured data, challenges include text mining and natural language processing (NLP). Text mining aims to uncover patterns and trends through statistical learning and text vectorization, supporting applications such as sentiment analysis and affective computing, which studies emotions in texts and human interactions. In this project, text mining is closely linked to affective computing and behavioral analysis, also encompassing image and video processing.

Behavioral analysis examines individuals within social networks, using graph-based models to identify communities and understand interaction dynamics. Applications include targeted marketing and information diffusion, providing insights into collective and emotional patterns within human interactions.

Faculty Members Involved:

• Eduardo Soares Ogasawara (coordinator) 
• Eduardo Bezerra da Silva 
• Gustavo Paiva Guedes e Silva 
• Jorge de Abreu Soares 
• Kele Teixeira Belloze

Machine Learning (ML) is a branch of Artificial Intelligence dedicated to developing new algorithms and methodologies capable of identifying patterns and making decisions without explicit programming. Beyond practical applications, progress in this field depends on creating novel theoretical and computational approaches that enhance the efficiency, interpretability, and generalization capacity of models.

This research project investigates advanced ML methods, spanning traditional techniques, such as deep neural networks and probabilistic models, to emerging approaches including self-supervised learning, generative models, federated learning, and reinforcement learning. Additionally, the project aims to improve strategies for explainability and interpretability to make models more transparent and trustworthy, especially in critical applications.

A second fundamental pillar of this project is Optimization, a field that integrates with ML to improve model performance and solve complex problems across different domains. The project focuses on the design and application of methods for solving problems using linear, nonlinear, integer, and mixed-integer programming (through exact and/or heuristic methods), as well as bio-inspired metaheuristics such as ant colony optimization, genetic algorithms, and particle swarm optimization. Optimization techniques are applied to tasks such as tuning machine learning model parameters, feature selection, and neural network architecture design.

Finally, Affective Computing explores how ML algorithms can interpret, process, and respond to human emotional states. This includes investigating new methods for fusing physiological and emotional signals. The goal is to advance the development of systems capable of adapting their responses in more natural and empathetic ways, with applications ranging from conversational interfaces to interactive robotics.

Faculty Members Involved: 

• Eduardo Bezerra da Silva (coordinator) 
• Gustavo Paiva Guedes e Silva 
• Diogo Silveira Mendonça 
• Diego Moreira de Araújo Carvalho 
• Laura Silva de Assis

The growing volume of data requires organizations to develop strategies for extracting valuable insights and gaining competitive advantage. This process involves the collection, storage, integration, and analysis of structured, semi-structured, and unstructured data. The research investigates methodologies for managing and transforming these data into useful knowledge to support decision-making.

The focus lies on data-centric artificial intelligence (Data-Centric AI) for data preparation and on large-scale processing techniques. One of the challenges addressed is the parallel and distributed processing of massive volumes of heterogeneous data, common in fields such as bioinformatics, astronomy, and engineering. Scientific workflows are essential for these experiments and are frequently executed on clusters, supercomputers, and cloud environments.

The project also explores frameworks such as Apache Spark, optimizing workflows for large-scale data analysis and management. In addition, it investigates conceptual modeling techniques, ontologies, preprocessing, indexing, and querying in Big Data systems. The research considers approaches based on distributed storage (HDFS), NoSQL databases, NewSQL systems, and object-relational databases, aiming to enhance the efficiency of data handling and analysis.

Faculty Members Involved:

• Rafaelli de Carvalho Coutinho (coordinator)  
• Eduardo Soares Ogasawara 
• Diego Moreira de Araújo Carvalho 
• Jorge de Abreu Soares 
• Kele Teixeira Belloze