https://tedebc.ufma.br/jspui/handle/tede/1314 Dissertações defendidas no âmbito do Programa. Sat, 23 May 2026 19:04:09 GMT 2026-05-23T19:04:09Z https://tedebc.ufma.br/jspui/handle/tede/7001 Título: Otimização Evolutiva Híbrida para Docking Proteína–Peptídeo Autor: RIBEIRO JUNIOR, Osmar Aguiar Primeiro orientador: CARMONA CORTÉS, Omar Andrés Abstract: Protein–peptide interactions play an important role in biological processes associated with molecular recognition, cellular signaling, and the development of therapeutic strategies. However, the computational prediction of these complexes remains challenging, especially in blind docking scenarios, in which the binding site is not previously known and the conformational space to be explored is broad. The central problem investigated in this work is to reduce the computational cost of the conformational search in blind protein–peptide docking without compromising the energetic quality of the final poses. In this context, this dissertation presents the development, implementation, and validation of a hybrid evolutionary optimization approach named ABC–GA–VGOS, applied to protein– peptide molecular docking. The proposed approach combines population-based exploration mechanisms, genetic operators, adaptive local refinement, and a single-objective geometric function based on KD-tree queries, used as a computationally efficient evaluation strategy during the search process. At the end of the optimization, the best pose obtained in each execution is submitted to an energy rescoring step inspired by AutoDock Vina, allowing the separation between fast geometric screening and final energetic evaluation. The experiments were performed with ten protein–peptide complexes obtained from the Protein Data Bank (PDB), considering binding energy, execution time, and comparative statistical analysis among the evaluated algorithms. The results indicated that the hybrid method obtained the lowest mean energies in five evaluated systems and achieved the shortest mean execution time in all cases, maintaining competitive behavior across different complexes. Thus, the main contribution of this work lies in the integration of complementary evolutionary strategies with efficient geometric evaluation, offering an alternative to reduce the computational cost of protein–peptide docking without relying exclusively on more expensive energy functions throughout the entire search. Instituição: Universidade Federal do Maranhão Tipo do documento: Dissertação Fri, 17 Apr 2026 00:00:00 GMT https://tedebc.ufma.br/jspui/handle/tede/7001 2026-04-17T00:00:00Z https://tedebc.ufma.br/jspui/handle/tede/6963 Título: Modelagem preditiva explicável para apoio à prevenção da evasão em cursos de graduação Autor: SILVA, Ronald César Carneiro da Primeiro orientador: CLÍMACO, Francisco Glaubos Nunes Abstract: Academic dropout remains a recurring challenge in higher education institutions, generating personal impacts for students as well as relevant social and institutional consequences. In recent years, machine learning models have been used to estimate the risk of academic dropout based on students’ academic, demographic, and socioeconomic data. However, despite their good predictive performance, many of these models operate with limited transparency, making it difficult to interpret their results. In this context, Explainable Artificial Intelligence (XAI) approaches emerge as an alternative to make predictions more understandable and practically useful. This dissertation investigates how Explainable Artificial Intelligence techniques can be applied to predictive models to estimate the risk of academic dropout in undergraduate programs, aiming to identify the factors that most influence model decisions and to analyze the potential of the generated explanations as support for planning institutional strategies to prevent dropout. The research was conducted using data from on-campus undergraduate programs at the Federal University of Maranhão (UFMA) and adopted a structured analytical approach that combines exploratory analysis, course-segmented predictive modeling, and global and local explainability techniques. Different machine learning algorithms were used to estimate dropout risk, and XAI methods such as SHAP, Permutation Importance, LIME, and Anchors were applied to interpret predictions and analyze the contribution of variables to model decisions. The results indicate that predictive models show good performance in identifying students at risk, although with variations across programs. The explanatory analyses revealed convergence among techniques in identifying three academic dimensions as central to the structure of predictions: efficiency in converting enrolled workload into course approvals (IECH), the pace of progression throughout the program (IEPL), and overall academic performance represented by the completion GPA (MC). In the analyzed context, demographic variables showed less significant influence on predictions. Furthermore, the results suggest that integrating predictive modeling with explainability techniques makes dropout risk more interpretable by highlighting academic performance patterns associated with the event. The explanations help clarify which factors sustain risk estimates across different programs. Consequently, this approach can support academic management in identifying student profiles that require greater attention and in planning institutional actions aimed at promoting student retention. Instituição: Universidade Federal do Maranhão Tipo do documento: Dissertação Wed, 13 May 2026 00:00:00 GMT https://tedebc.ufma.br/jspui/handle/tede/6963 2026-05-13T00:00:00Z https://tedebc.ufma.br/jspui/handle/tede/6962 Título: Segmentação automática do pâncreas em tomografias computadorizadas abdominais utilizando uma abordagem orientada a atlas probabilístico e aprendizado profundo Autor: TELES, Felipe Rogério Silva Primeiro orientador: DINIZ, João Otávio Bandeira Abstract: Automatic segmentation of the pancreas in Computed Tomography (CT) images is a challenging task due to the organ’s high anatomical variability, low contrast between adjacent tissues, and the large amount of data present in volumetric examinations. This task is important for various clinical applications, such as computer-aided diagnosis, treatment planning, and quantitative analysis of anatomical structures. However, manual identification of the pancreas in CT scans can demand significant time from specialists, making the process susceptible to variations between observers. In this context, computational methods based on deep learning have been widely used to assist in the automatic analysis of medical images. Thus, this work proposes an automated method for pancreas segmentation in abdominal CT images. The developed method consists of different stages, including image preprocessing, automatic filtering of slices without the organ using a convolutional neural network, extraction of the region of interest using a probabilistic atlas, and application of deep learning models to perform the segmentation. For this stage, different neural network architectures were trained, including DeepLabV3, U-Net++, and SegFormer. Subsequently, the results generated by the models were combined using an ensemble strategy to increase the robustness of the segmentation. The method was evaluated using the Medical Segmentation Decathlon database, composed of abdominal CT scans manually annotated by specialists. The experimental results demonstrated that the proposed ensemble achieved superior performance to the individual models, obtaining a Dice coefficient of 78.55%, an IoU of 68.17%, and a Recall of 84.41%. Thus, the results obtained indicate that the proposed approach has the potential to assist specialists in the analysis of CT scans, contributing to the development of diagnostic support systems based on deep learning. Instituição: Universidade Federal do Maranhão Tipo do documento: Dissertação Mon, 27 Apr 2026 00:00:00 GMT https://tedebc.ufma.br/jspui/handle/tede/6962 2026-04-27T00:00:00Z https://tedebc.ufma.br/jspui/handle/tede/6960 Título: Treinamento de funções de ativação em redes neurais artificiais: uma abordagem com regressão polinomial Autor: CAJADO, Carlos Eduardo Nascimento Primeiro orientador: ALMEIDA NETO, Areolino de Abstract: Artificial neural networks are widely applied to complex problems; however, selecting an optimal activation function (AF) remains a challenge, particularly due to the lack of consistent criteria for relating data characteristics to the most appropriate function. In light of these limitations, recent advances highlight the potential of trainable activation functions, which dynamically adjust during training to improve network performance. Such a mechanism can also be leveraged to insert new hidden layers into stacked autoencoder networks, promoting greater structural depth without compromising training stability. In this context, this work proposes a novel approach based on polynomial regression to develop trainable activation functions. The methodology introduces a collaborative layer insertion mechanism, enabling the integration of new layers without degrading the knowledge acquired in previous ones, while simultaneously enhancing the network’s adaptive capacity and performance through the use of error estimation as a criterion for dynamically adjusting the outputs of hidden neurons, in comparison with fixed activation functions. The proposed approach is evaluated on eight benchmarking datasets extracted from the OpenML platform, totaling 4,800 experiments, analyzing its impact on training stability, adaptability, and overall network performance compared to the traditional sigmoid activation function. Instituição: Universidade Federal do Maranhão Tipo do documento: Dissertação Fri, 10 Apr 2026 00:00:00 GMT https://tedebc.ufma.br/jspui/handle/tede/6960 2026-04-10T00:00:00Z