http://bibliotecatede.uninove.br/handle/tede/2569 2026-09-02T11:10:07Z 2026-09-02T11:10:07Z Doretto, Matheus Martins http://bibliotecatede.uninove.br/handle/tede/4014 2026-08-27T15:33:50Z 2026-06-17T00:00:00Z

Título: Recursos educacionais baseados em visão computacional para analisar a relação entre atenção e desempenho escolar Autor: Doretto, Matheus Martins Primeiro orientador: Pinto, Luiz Fernando Rodrigues Abstract: Monitoring attention in the classroom represents a recurring challenge in the educational context, especially because traditional methods, such as direct observation, tend to be subjective and poorly suited to systematically tracking students’ attentional variations. In this context, computer vision emerges as a promising alternative for the automated extraction of visual information associated with student behavior. Based on the literature review, a scarcity of studies was identified that directly relate attention measured through computer vision to school performance. Therefore, the objective of this research was to develop and apply a computational method to correlate attention indicators identified in classroom images with students’ school performance. To this end, a mixed-method approach was adopted, with an applied, exploratory, correlational, and documentary nature, based on data provided by a private educational institution in the city of São Paulo. Classroom images were processed using the YOLOv11 and YOLO Face models, enabling the detection of people and faces and the extraction of data used in the heuristic classification of attention levels. Subsequently, the frequencies of these levels were consolidated and converted into attention scores on a scale from 0 to 10 and organized with school performance scores, forming data pairs for descriptive and correlational analyses. The results showed greater stability in attention scores compared to performance scores, as well as a positive association between the variables. The Pearson coefficient was r = 0.285, with p = 0.127, indicating a weak and non-significant positive correlation. The Spearman coefficient was ρ = 0.375, with p = 0.041, indicating a moderate and statistically significant positive correlation. It is concluded that the proposed approach demonstrated feasibility in transforming image records into quantitative attention indicators and correlating them with school data. As a contribution, the study presents a methodological path applicable to educational analysis, although limited by the sample scope and the adopted heuristic, indicating the need for future research with expanded datasets and new visual indicators. Instituição: Universidade Nove de Julho Tipo do documento: Dissertação

2026-06-17T00:00:00Z Izidoro, Magdiel de Brito http://bibliotecatede.uninove.br/handle/tede/4013 2026-08-27T15:19:34Z 2026-04-27T00:00:00Z

Título: Uma análise de redes complexas da mobilidade urbana ferroviária de São Paulo a partir de dados de bilhetagem eletrônica de 2019 e 2020 Autor: Izidoro, Magdiel de Brito Primeiro orientador: Schimit, Pedro Henrique Triguis Abstract: This study analyses rail-based urban mobility in the São Paulo Metropolitan Region using anonymized electronic ticketing records from 2019 and 2020. The study addresses a common limitation of smart card data, which generally records only passenger boarding events without explicit information on alighting locations. To overcome this limitation, an origin-destination inference procedure based on the Transaction Sequence Number (TSN) is proposed, allowing the reconstruction of successive trips and the construction of daily mobility networks. Based on these networks, structural metrics from complex network theory were analysed, together with distance distributions, station-level entropy, ranking dynamics, and spatial comparisons between periods. The results show that the COVID-19 pandemic was associated with a strong contraction in demand and significant changes in the structural organization of the network. A reduction of 41.2% in the total volume of trips was observed, accompanied by a 44.2% decrease in average network strength and a 44.0% reduction in global efficiency. Conversely, the average path length increased by 64.0%, together with a 9.0% growth in the number of communities and a spatial reorganization of flows between stations and system corridors. These findings indicate that the pandemic affected not only travel demand but also the way rail mobility was organized across the metropolis. Therefore, this study demonstrates that electronic ticketing data, when combined with origin-destination inference methods and complex network analysis, can contribute to the continuous monitoring of urban mobility and support the planning and management of public transportation systems. Instituição: Universidade Nove de Julho Tipo do documento: Dissertação

2026-04-27T00:00:00Z Gomes, Fernanda Pereira http://bibliotecatede.uninove.br/handle/tede/4012 2026-08-27T15:13:14Z 2026-06-03T00:00:00Z

Título: Mineração de dados educacionais na identificação do perfil de alunos com risco de evasão em escola técnica pública com base na trajetória acadêmica Autor: Gomes, Fernanda Pereira Primeiro orientador: Sassi, Renato José Abstract: Public Technical Schools in the State of São Paulo, known as ETECs, seek to align their educational programs with labor market demands through courses that combine theory and practice. These institutions face challenges related to school dropout, which can be examined through institutional academic data. The case of ETEC Raposo Tavares illustrates this context: only 43.59% of the students who entered the Technical Course in Chemistry in 2023 completed the program in 2024, a percentage below the institutional target of 65%. In this context, Educational Data Mining (EDM) makes it possible to explore these data and identify patterns associated with students at risk of dropout. This study aims to apply Educational Data Mining to identify profiles of students at risk of dropout in the Technical Course in Chemistry at ETEC Raposo Tavares, based on their academic trajectory, and to support school management in the development of educational strategies. The database used in the computational experiments contains information on 185 students enrolled from the first to the third module in 2023 and 2024. The data were extracted from the New Academic System (NSA) and supplemented with records from class council minutes. The experiments followed the phases of the Knowledge Discovery in Databases process: Data Selection, Data Preprocessing, Data Transformation, Educational Data Mining, and Knowledge Interpretation and Evaluation. Logistic Regression, Random Forest, and XGBoost were applied, and Random Forest achieved the best performance. Based on its results, the school management team identified three profiles: High Academic Performance, associated with students who are committed to the course and seek completion; Medium Academic Performance, associated with students who, despite difficulties, seek to complete the course; and Low Academic Performance, associated with a higher risk of dropout. The last profile may be related to social inequalities, lack of family support, vocational misalignment, or a weak sense of institutional identification and belonging. Among the measures selected by the management team, the institutionalization of an internal continuous monitoring protocol stands out. The protocol should connect the indicators to pedagogical practices so that the data can support concrete educational strategies. The study concludes that EDM is a relevant resource for ETEC Raposo Tavares and can support school management in the development of strategies aimed at reducing the risk of dropout. Instituição: Universidade Nove de Julho Tipo do documento: Dissertação

2026-06-03T00:00:00Z Oliveira, Reinaldo Ribeiro de http://bibliotecatede.uninove.br/handle/tede/3986 2026-05-26T15:29:36Z 2026-03-10T00:00:00Z

Título: Novo índice clínico-laboratorial para predição de deterioração em adultos hospitalizados com modelo híbrido de aprendizagem profunda Autor: Oliveira, Reinaldo Ribeiro de Primeiro orientador: Dias, Cleber Gustavo Abstract: INTRODUCTION: Clinical deterioration in patients, which may lead to increased mortality risk, particularly in intensive care units, is associated with a dynamic process of physiological decline characterized by the progression of organ dysfunction. This process can be monitored over time through measurable indicators obtained from patients’ clinical records. Assessing clinical deterioration is a demanding task due to the complex and variable clinical behavior of each patient. The growing volume of clinical data within Electronic Health Records (EHRs) is characterized as longitudinal patient health information. These data enable the application of clinical deterioration protocols capable of observing and tracking physiological progression in hospitalized patients. The use of EHRs allows electronic records to generate insights and predictions of future clinical events. This study is justified by the need to develop a novel index grounded in hybrid artificial intelligence capable of processing heterogeneous temporal clinical data derived from electronic health records. OBJECTIVE: To develop and test an index composed of clinical and laboratory data for predicting deterioration in hospitalized adult patients using machine learning algorithms. METHODS: This study has an exploratory experimental technological design. A total of 1,100,000 patients were selected from two datasets originally extracted from the MIMIC-III database. The study was structured into three stages: (i) a scoping review, (ii) development of an adult patient deterioration index based on the most relevant clinical and laboratory time-series data treated statistically, and (iii) computational experiments using a hybrid approach combining convolutional neural networks (CNNs) and long short-term memory (LSTM) networks to predict the deterioration index at future time steps. Mathematical models were constructed to calculate the weights of variables composing the deterioration index. RESULTS: The hybrid neural network model was evaluated across 10 different scenarios with hyperparameter tuning (Seed, Past_H, Future_H, Val_Split, Epochs, and Batch_Size). The model demonstrated the ability to predict future events using retrospective time-series records to forecast 6-hour, 12-hour, 18-hour, and 24-hour horizons. Global performance metrics achieved were MAE ranging from 9.36 to 9.89 and MAPE from 24.19% to 25.96%. CONCLUSIONS: The hybrid model showed consistent performance under similar hyperparameters across different patient cohorts from MIMIC-III and MIMIC-IV. The best results were observed for 6-hour-ahead predictions. The study presents innovative potential and offers scientific contributions to the academic community, society, and healthcare delivery processes. The findings suggest that the proposed approach can anticipate clinical events and provide opportunities for timely interventions and clinical decision-making in hospital settings, thereby enhancing patient safety. Instituição: Universidade Nove de Julho Tipo do documento: Tese

2026-03-10T00:00:00Z