http://repositorio.ufc.br/handle/riufc/24011 2026-04-06T08:08:05Z 2026-04-06T08:08:05Z Freitas, Leyverson de Oliveira http://repositorio.ufc.br/handle/riufc/85289 2026-03-10T16:13:41Z 2026-01-01T00:00:00Z

Título: Conformidade do Portal gov.br à WCAG 2.2: uma análise comparativa com WAVE e Accessibility Insights for Web Autor(es): Freitas, Leyverson de Oliveira Abstract: Web accessibility aims to ensure that people with different types of disabilities can perceive, understand, navigate, and interact with digital content and services in an autonomous and effective way. This study aims to evaluate the accessibility of the GOV.BR portal based on the Web Content Accessibility Guidelines (WCAG 2.2), using automated evaluation tools. The research seeks to identify barriers that compromise the browsing experience of users with disabilities and to propose recommendations that contribute to improving the portal. To achieve this objective, five representative pages of the website were analyzed using the tools WAVE Web Accessibility Evaluation Tool and Accessibility Insights for Web. The analyses revealed issues such as low contrast between text and background, links and buttons without adequate descriptions, form fields without accessible labels, incorrect semantic structure, and absence of alternative text in images. Based on these results, recommendations were developed to address these nonconformities, aiming to increase the portal’s level of compliance with WCAG guidelines and to promote a more inclusive experience for all users. The study reinforces the importance of digital accessibility as a continuous commitment to inclusion and equal access to information. Tipo: TCC

2026-01-01T00:00:00Z Lima, Henrique Lopes http://repositorio.ufc.br/handle/riufc/85088 2026-03-10T16:12:05Z 2026-01-01T00:00:00Z

Título: Implementação e análise de algoritmos de caminho mínimo para cálculo de matrizes de distância Autor(es): Lima, Henrique Lopes Abstract: This work presents the theoretical foundation and methodology for implementing the Uniform Cost Search algorithm for computing shortest paths in directed, weighted graphs with nonnegative weights. The principles of the algorithm, its pseudocode implementation, and a methodology for performance evaluation are discussed, focusing on parallel implementations in Go and C++. The analysis covers time complexity, memory usage, and suitability for the problem, providing a robust foundation for practical and experimental studies on large-scale graphs. Tipo: TCC

2026-01-01T00:00:00Z Nascimento Júnior, José Fábio dos Santos http://repositorio.ufc.br/handle/riufc/85010 2026-02-27T17:43:31Z 2026-01-01T00:00:00Z

Título: Valiação Funcional da interpretabilidade de modelos de classificação Boosting para previsão de inadimplência de crédito Autor(es): Nascimento Júnior, José Fábio dos Santos Abstract: Credit granting plays a central role in the economy, requiring financial institutions to strike a delicate balance between accurately detecting default and ensuring transparency in automated decisions. This study investigates the application of ensemble classifiers, with a focus on boosting techniques, for predicting credit default. Given the importance of the topic and the challenges of reconciling high predictive performance with interpretability, the study proposes an integrated analysis of agnostic explainability methods, with an emphasis on the application of SHAP and LIME frameworks for individual prediction auditing. The models are quantitatively evaluated for fidelity, robustness, attribute importance, and representativeness of explanations, with the aim of providing support for the safe and transparent adoption of these solutions in regulated contexts. The experimental results highlighted LightGBM as the reference architecture in discrimination, achieving the largest area under the ROC curve of 0.7846, while XGBoost consolidated itself for operational efficiency, delivering the highest accuracy and balance by the F1-Score. The analysis revealed that variables from external sources exert a dominant influence on the decision, followed by demographic factors. High stability was observed in the justifications, fundamental for consistency in communication with the client, despite the non-linear complexity evidenced by the low fidelity of local approximations. Thus, the research presents an effective strategy for modernizing credit analysis, ensuring that the gain in predictive performance is accompanied by the transparency indispensable to regulatory compliance. Tipo: TCC

2026-01-01T00:00:00Z Carvalho, Yander Alves de http://repositorio.ufc.br/handle/riufc/85008 2026-02-27T17:23:33Z 2026-01-01T00:00:00Z

Título: Avaliação de usabilidade de um chatbot educacional na plataforma do Discord Autor(es): Carvalho, Yander Alves de Abstract: This work explores programming education and the emergence of educational chatbots as tools for more dynamic learning. The COVID-19 pandemic accelerated reliance on the digital environment, highlighting the need to integrate pedagogical tools into communication platforms already used by students. The general objective is to perform an exploratory usability evaluation of the ’Hora de Codar’ chatbot in a controlled environment to identify interface barriers and opportunities for improvement in the technical interaction flow. The methodology adopted a mixed approach, including the definition of practical use scenarios (registration, progress verification, activity query and submission, and ranking), user profile collection, and usability evaluation via the System Usability Scale (SUS). The analysis combined quantitative results from the SUS with an investigation of interaction logs, focusing on identifying barriers in command-line interaction. The conclusions point to a high acceptance of the tool but highlight specific interface challenges, guiding improvement proposals such as confirmation mechanisms and the use of native platform commands Tipo: TCC

2026-01-01T00:00:00Z