http://repositorio.ufc.br/handle/riufc/384 2026-04-14T17:19:36Z 2026-04-14T17:19:36Z Silva Júnior, Lourival Gerardo da http://repositorio.ufc.br/handle/riufc/84824 2026-02-19T12:23:25Z 2025-01-01T00:00:00Z

Título: SEPTRON : Predição de sepse usando transformer, ontologia e algoritmo bioinspirado Autor(es): Silva Júnior, Lourival Gerardo da Abstract: The use of Machine Learning (ML) techniques for sepsis prediction has evolved globally, but the selection of clinically relevant variables, essential for building robust predictive models, remains one of the main challenges. This aspect is particularly important for patients with suspected sepsis, who require intensive monitoring during the first hours after admission to Intensive Care Units (ICUs), in order to enable early diagnosis and treatment. In this context, this work proposes the model SEpsis Prediction with TRansformer, Ontology and BioiNspired Algorithm (SEPTRON), built on a Transformer architecture adapted for structured data. The model uses variables initially selected from a conceptual ontological model of sepsis and further refined by a Genetic Algorithm (GA). The contributions include: (i) a structured and replicable pipeline, covering the entire process from data extraction to model evaluation; (ii) creation of the curated dataset MIMIC-IV-EXT-ONTO-SEPSE, derived from the MIMIC-IV database; (iii) Transformer architecture adapted for structured data; and (iv) evidence of the relevance of the chosen clinical variables. The comparative evaluation focused on the performance of the Transformer against four classic models: Logistic Regression, Random Forest, XGBoost, and Long Short-Term Memory (LSTM). Using 4-hour time windows over the first 72 hours of ICU stay, the Transformer with all 17 clinical variables achieved superior performance (AUC-ROC 0.85; recall 0.83). In turn, the SEPTRON, which integrates the Genetic Algorithm to reduce the set of clinical variables from 17 to 8, maintained competitive performance (AUC-ROC 0.83; recall 0.76), combining robustness, interpretability, and computational feasibility. This effectiveness was reinforced by the temporal analysis of the evolution of sepsis probability over the 72 hours. It is concluded, therefore, that SEPTRON represents a relevant proposal to the field of digital health, providing a predictive model capable of anticipating the risk of sepsis in ICU with a good balance between performance, interpretability, and clinical applicability. Tipo: Tese

2025-01-01T00:00:00Z Pereira, Danne Makleyston Gomes http://repositorio.ufc.br/handle/riufc/84292 2026-01-16T16:28:38Z 2025-01-01T00:00:00Z

Título: A microservices-based software architecture for building flexible smart city platforms Autor(es): Pereira, Danne Makleyston Gomes Abstract: A smart city integrates data from various subdomains to offer intelligent services, improving city resource management and citizens’ daily lives. Technologies such as ICT, IoT, big data, and AI enable this data management, which is supported by software platforms. However, in addition to the usual challenges of software platforms, developing a smart city platform faces issues such as resource heterogeneity, flexibility, AI efficiency, fast data processing, and low latency. This work proposes a microservice-oriented, data-centric platform with a three-layer architecture (edge, fog, cloud) to address these issues. This approach improves flexibility, scalability, and latency management, ensuring a responsive and efficient solution for smart cities. Semantic annotation mechanisms in the fog layer resolve data heterogeneity, enabling data exchanges using semantic values. This work’s main contributions include identifying essential functionalities for smart city platforms, analyzing existing solutions, specifying a reference architecture, implementing a prototype, and conducting performance evaluations. Initially, we conducted a Systematic Literature Review (SLR) to understand the state of the art, identifying the current enabling technologies, essential software platform requirements, and relevant open issues. Based on this, we introduce UFCity, a software architecture designed to meet such demands. We analyzed a UFCity-based prototype both qualitatively (use case scenarios) and quantitatively (Experimental Design 33). In the use case scenarios, we observed a reduction in the number of message exchanges and network bandwidth usage due to distributed data processing across solution layers. Furthermore, semantic mechanisms and IoT middleware efficiently handled resource heterogeneity. In these use case scenarios, we identified several functional requirements listed in this work. In quantitative analysis, we found consistent throughput in fog computing nodes even with message overload. Thus, this work demonstrates that UFCity meets the essential requirements of a smart city platform and exhibits high execution performance, establishing itself as an advantageous olution compared to those proposed in the literature. Tipo: Tese

2025-01-01T00:00:00Z Figueredo, Pedro Jorge de Abreu http://repositorio.ufc.br/handle/riufc/83956 2025-12-19T18:52:08Z 2025-01-01T00:00:00Z

Título: O problema de alocação de pontos de controle Autor(es): Figueredo, Pedro Jorge de Abreu Abstract: We formally introduce the Multi-period Checkpoint Allocation Problem with Conflicts (PAPCMC), which allocates indivisible, mobile control units to discrete locations over a finite time horizon, maximizing the total weight of allocations under exclusivity, compatibility, and movement constraints. We show that PAPCMC can be reduced to an Assignment Problem with Conflicts (APC). Additionally, we prove that the problem is NP-hard even with only two periods, via a polynomial reduction from the (3,B2)-SAT problem. We present two alternative integer linear programming formulations for the problem: BLP1, which encodes a solution as a maximum independent set, and BLP2, which represents PAPCMC as a maximum-cost flow problem with conflict constraints. We show that BLP2 has a linear relaxation at least as tight as that of BLP1 and can be solved in polynomial time when all teams share identical schedules. Moreover, its block structure enables decomposition into per-team subproblems, inspiring efficient heuristics. We designed three main heuristics—HGC, HGE, and HL—as well as versions inspired by methods GRASP and ACO. We created a set of 15198 realistic instances based on public data from New York City. In tests with a 2-minute time limit, HGC outperformed HGE in 96.3% of the instances in terms of objective value; HL reached optimality in more than half of the instances and reduced the average gap to 1%. Incorporating the heuristics’ lower bounds proved effective. Specifically, BLP2 achieved optimal solutions in 94.1% of the instances, reduced average computation time by 40%, and decreased the integrality gap by 35% relative to BLP1. Even with a 1-hour execution limit on difficult cases, BLP2 maintained superiority in the number of optimal solutions found. We conclude that integrating lower bounds from heuristics reduces total computation time and increases the number of instances solved to optimality: adopt HGC when the goal is to find an optimal solution and HL for stronger lower-bound guarantees. Given its overall performance, BLP2 is the recommended approach in operational scenarios. Tipo: Tese

2025-01-01T00:00:00Z Barbosa, Pedro Luís Saraiva http://repositorio.ufc.br/handle/riufc/83503 2025-11-21T21:01:25Z 2025-01-01T00:00:00Z

Título: Computerized adaptive testing as a tool for assessment and preparation for Enade Autor(es): Barbosa, Pedro Luís Saraiva Abstract: This thesis investigated the impact of using Computerized Adaptive Testing (CAT) on the perception and acceptance of students in Computing Education programs regarding their preparation for the National Student Performance Exam (Enade). The research stems from the observation that many students face difficulties in engagement and motivation during exam preparation, as well as a possible disconnection between classroom teaching and the assessment format. Considering the strategic role of Enade within the National Higher Education Evaluation System (SINAES), the central problem addressed is the search for technological strategies that make the exam preparation process more personalized, motivating, and effective. To address this issue, the study proposed and developed a solution based on an adaptive model that integrates CAT logic with the Elo rating system, implemented in the Questione platform. The investigation followed the principles of the Design Science Research Methodology (DSRM), encompassing the stages from problem identification to the development, demonstration, and evaluation of the artifact. The proposed model was validated through an empirical study that compared student performance and perception in fixed and adaptive mock exams, using the Technology Acceptance Model (TAM) as the theoretical basis for acceptance analysis. The results showed that, although the average performance was slightly lower in adaptive mock exams (due to the higher level of question difficulty) students demonstrated greater acceptance, with significant improvements in the dimensions of perceived usefulness, attitude, and intention to use. The study concludes that the adaptive model positively impacts student acceptance and represents an alternative for personalizing Enade preparation, contributing both methodologically and technologically to the field of large-scale adaptive assessment. Future directions include the automated expansion of the question bank supported by language models, the use of multidimensional adaptive models, and the enhancement of accessibility in adaptive educational contexts. Tipo: Tese

2025-01-01T00:00:00Z