Transfer learning models for image falsification detection in faces

Rostos, Modelos de Transfer Learning para Detecção de Falsificação de Imagens em

doi:10.22167/2675-6528-202602724

Article

September 30, 2026

Transfer learning models for image falsification detection in faces

Transfer Learning Models for Face Image Forgery Detection

Bianca Cavalcante Basso; Jorge Costa Silva Filho

DOI: 10.22167/2675-6528-202602724

Article derived from a Course Conclusion Work (TCC), with content based on the student’s original work and adapted to the editorial format of the E&S Magazine with the support of the ResumeAI tool, an artificial intelligence solution developed by Instituto Pecege for textual synthesis and organization.

Summary

The advancement of Artificial Intelligence (AI) has consolidated its application in the generation of synthetic media, such as deepfakes, frequently employed in disinformation and fraudulent practices. This work evaluated the effectiveness of pre-trained transfer learning models in distinguishing between real and synthetic images, specifically deepfakes. For this purpose, the Celeb-DF v2 dataset, composed of real and synthetic videos, was used. Preprocessing was performed for frame extraction and image resizing, followed by downsampling to balance the classes, totaling 18,358 images. The ResNet50, VGG16, and MobileNetV2 models were trained for 10 epochs, with their last convolutional layers fine-tuned and the first ones frozen, and subjected to internal validation tests and an independent test set. Internal validation results indicated high performance, with ResNet50 achieving an AUC of 0.8762, MobileNetV2 0.8279, and VGG16 0.7625. However, in independent tests, the models presented a generalization gap, with accuracies stabilizing close to 50%. It was concluded that, although data rebalancing and transfer learning are effective for learning specific features, robust deepfake detection in real-world scenarios requires models that overcome texture dependency and generalize synthesis patterns, going beyond nominal performance in controlled environments.

Keywords: Computer vision; Deepfake; Artificial intelligence; Machine learning; Transfer learning.

1. Introduction

Artificial Intelligence (AI) has been consolidating in various sectors, including the generation of synthetic media such as images, videos, and audios (MONTAÑO, 2023). In this scenario, deepfakes emerge as media generated or altered by deep learning algorithms that simulate real content.

Despite its potential for legitimate uses, this technology is frequently employed in the propagation of disinformation and fraudulent practices (KAUFMAN, 2022). Brazil, for example, ranks third globally in social media consumption (Forbes, 2023), which intensifies the challenges related to verifying the authenticity of digital content and favors the dissemination of deepfakes (JUNIOR, 2023).

Deepfakes are created through deep learning techniques, a branch of AI that relies on multi-layered neural networks (WESTERLUND, 2019). Among these, Convolutional Neural Networks (CNNs) stand out, widely used in image processing. CNNs operate with layers such as Convolution, Rectified Linear Units (ReLU), and Max-Pooling, which extract visual features and reduce data dimensionality (SILVA, 2018; JUNIOR, 2023).

The versatility of these architectures allows for digital manipulations with a high degree of realism, making it difficult to distinguish between real and fake content (WESTERLUND, 2019). Although there are beneficial applications, the use of deepfakes has been diverted to the dissemination of disinformation, generating negative social impacts (JUNIOR, 2023). The production of media of dubious provenance, combined with the speed of sharing, consolidates mistaken perceptions of veracity in the public. A notorious example was the manipulation of a video involving Dr. Drauzio Varella for fraudulent promotion, evidencing risks to moral integrity and reputation (MACHADO, 2023; JUNIOR, 2023).

The availability of free tools for algorithm development, combined with the practicality of Transfer Learning, has democratized access to Artificial Intelligence. However, this accessibility also acts as a catalyst for the increase of deepfakes, as individuals with malicious intentions can employ these resources to create disinformation (JUNIOR, 2023). The literature suggests that the advancement of AI should be accompanied by ethical principles and meticulous inspection of the generated results, aiming to mitigate the risks of disinformation (KAUFMAN, 2022).

Given this scenario, the effective detection of deepfakes becomes crucial for the security and authenticity of digital information. The complexity of high-fidelity manipulations demands robust models capable of distinguishing between real and synthetic media, overcoming generalization challenges. Therefore, this case study aims to compare the performance of three pre-trained models (Transfer Learning), namely ResNet50, VGG16, and MobileNetV2, in identifying deepfakes, using a dataset composed of real and synthetic media.

2. Material and Methods

The present study adopted a quantitative and experimental methodological approach, focused on the implementation and evaluation of machine learning models for deepfake detection. The research consisted of applying Transfer Learning techniques, using pre-trained convolutional neural network architectures to classify images as real or synthetic. The objective was to compare the performance of these models in identifying digital manipulations in faces.

For the development, the Celeb-DF v2 dataset (A Large-scale Challenging Dataset for DeepFake Forensics) was employed, recognized as a high-quality benchmark (LI et al., 2020). This set comprises 5,639 AI-manipulated videos and 590 original videos, collected from the YouTube platform, presenting diversity in ethnicity, age, and gender.

The preprocessing involved extracting frames from videos, converting them into static data for analysis by computer vision models. The MTCNN library was used for facial detection and OpenCV (cv2) for frame manipulation. Sampling consisted of extracting one frame every fifteen seconds, with a margin factor of 20% on the facial crop.

The images were resized to 224 x 224 pixels, a standard compatible with Transfer Learning architectures, converted to RGB, and exported as .jpg. The final dataset, after preprocessing, totaled 98,482 images (9,179 “real” and 89,303 “synthetic”).

To mitigate the class imbalance, the downsampling technique was applied to the majority class. This process resulted in a balanced training set of 18,358 images, with a 1:1 ratio between classes. `shuffle` was applied immediately after downsampling to prevent the model from learning patterns based on the file reading order.

The development environment was the Google Colab platform, operating in a Jupyter Notebook. The cloud-based infrastructure provided graphical processing unit (GPU) and tensor processing unit (TPU) resources, essential for training deep neural networks. File storage and management were carried out via integration with Google Drive.

For the modeling, three pre-trained Convolutional Neural Network (CNN) architectures were selected for Transfer Learning: ResNet50, VGG16, and MobileNetV2. These architectures were chosen for their representativeness and robustness in computer vision tasks (SIMONYAN; ZISSERMAN, 2014; HE et al., 2016; SANDLER et al., 2018). VGG16 evaluated classic models, ResNet50 tested the ability to capture subtle details, and MobileNetV2 verified feasibility in resource-limited scenarios.

The training strategy consisted of freezing the initial convolutional layers of the models (`trainable = False`), leveraging the already optimized low-level feature extractors. The original output layer was replaced by a new binary classification structure, composed of GlobalAveragePooling2D, a Dense layer with 128 neurons and ReLU activation, and a Dropout layer with a rate of 0.5 for regularization. The final output layer used one neuron with a Sigmoid activation function.

The model compilation employed the Adam optimizer and the Binary Crossentropy loss function. After rebalancing, the dataset was split into 80% for training and 20% for validation and testing. Training was performed for 10 epochs. To ensure reproducibility and balanced class distribution in each training batch, the `image_dataset_from_directory` class from TensorFlow was used, configured with a batch size of 32 and shuffle enabled.

To assess the effectiveness of the models, statistical metrics based on the Confusion Matrix were adopted. The metrics included Accuracy, Precision, Recall (Sensitivity), F1-Score, ROC Curve, and Area Under the Curve (AUC), which measured the discrimination capacity between real and synthetic classes (PRATI; BATISTA; MONARD, 2008). The Confusion Matrix was used to visualize the distribution of hits and misses (TP, TN, FP, FN).

3. Results and Discussion

The results obtained in this study reveal a complex panorama regarding the effectiveness of transfer learning models in deepfake detection, evidencing a notable contrast between performance in internal validation and generalization capability in independent tests. Initially, the ResNet50, VGG16, and MobileNetV2 models demonstrated high performance in a controlled environment, after a rigorous process of preprocessing and rebalancing the Celeb-DF v2 dataset. However, when subjected to a more challenging test set, the models presented a significant drop in accuracy, stabilizing around 50%, which points to a crucial generalization gap for the discussion of the robustness of these systems in real-world scenarios.

Data preparation was a fundamental step to mitigate algorithmic bias and optimize the dataset. The MTCNN library was used for facial detection and OpenCV for frame manipulation, extracting one frame every interval proportional to the average duration of the videos, with a 20% margin to preserve facial features. The images were resized to 224 x 224 pixels, a standard compatible with the selected computer vision architectures. In the end, the main dataset contained 98,482 images, with 9,179 real and 89,303 synthetic.

To address the significant imbalance between the “Real” and “Fake” classes in the original dataset, the downsampling technique was applied. This process selectively reduced samples from the dominant class until both categories reached a 1:1 proportion, resulting in a balanced set of 18,358 images for training. This strategy was crucial to ensure that accuracy and precision metrics reflected the models’ true distinguishing capability, rather than just a statistical trend of the dataset, forcing the networks to learn real morphological patterns and manipulation artifacts.

The training strategy involved freezing the initial convolutional layers of the pre-trained ImageNet models (trainable = False), leveraging their optimized weights for low-level feature extraction. The original output layer was replaced by a new binary classification structure, consisting of GlobalAveragePooling2D for dimensionality reduction, a Dense Layer with 128 neurons and ReLU activation function, Dropout of 0.5 for regularization, and an output layer with 1 neuron and Sigmoid activation function. The compilation used the Adam optimizer with a learning rate of 0.001 and the Binary Crossentropy loss function, with the dataset split into 80% for training and 20% for validation and testing.

Performance of Architectures

After applying the downsampling technique and training for 10 epochs, the models exhibited significantly more robust predictive behavior in the internal validation phase. The ResNet50 architecture stood out with the best absolute performance, achieving an AUC of 0.8762. This value indicates a high capacity for correct distinction between real and synthetic classes, surpassing the other evaluated architectures. ResNet50 also obtained an accuracy of 0.7768, precision of 0.8320, recall of 0.7109, and F1-Score of 0.7667, with 506 True Negatives and 526 False Negatives.

The MobileNetV2 architecture, although with a slightly lower AUC (0.8279) than ResNet50, demonstrated a better balance between Precision (0.7443) and Recall (0.7420). This balance suggests that the model is more consistent in handling False Positive and False Negative errors, maintaining stable performance on both fronts. MobileNetV2 achieved an Accuracy of 0.7362, precision of 0.7443, recall of 0.7420, and F1-Score of 0.7431, with 470 True Positives and 478 False Negatives, indicating a higher sensitivity in identifying manipulated content.

In contrast, the VGG16 model presented the most modest performance among the three architectures, with an AUC of 0.7625 and an F1-Score of 0.6901. Its accuracy (0.6907), precision (0.7160), and recall (0.6659) rates were the lowest, suggesting that its structure may require a larger number of epochs or more refined hyperparameter adjustments for this specific dataset. The comparison between the preliminary and current results confirmed that the suppression of the majority class bias was decisive for the stabilization of metrics around 70-80% in accuracy and AUC, validating the effectiveness of Transfer Learning for the task of facial digital forensics.

The evolution of the ResNet50 model’s performance metrics during training, over the 10 epochs, demonstrated progressive convergence. The AUC index, for example, jumped from 0.59 in the first epoch to 0.86 in the tenth epoch. This continuous improvement indicates that the network was able to extract relevant features from the provided sample, learning to distinguish patterns more effectively as training progressed. Intermediate results in the fifth epoch already showed an AUC of 0.8202, confirming the model’s learning capability.

Generalization Gap Analysis

Despite the promising performance in internal validation, submitting the models to an independent test set revealed a sharp drop in performance, with metrics stabilizing around 0.50. This phenomenon, known as generalization gap, suggests a limitation in the models’ ability to apply knowledge acquired on test data that exhibits variations not observed during training. The test accuracy of ResNet50 was approximately 50.4% (920 correct predictions out of 1824 samples).

For the ResNet50 model, the analysis of the confusion matrix on the independent test indicated 506 True Negatives (real images correctly classified) and 526 False Negatives (deepfakes classified as real). The latter is the most critical error in a security scenario, as it means fraudulent media “passed” through the detection filter. The model demonstrated a conservative bias, tending to classify images as real, which can be problematic for threat detection.

The VGG16 model presented a test accuracy of approximately 50.7% (925 correct predictions out of 1824 samples) on the independent dataset. Its confusion matrix showed 469 True Negatives (Real) and 456 True Positives (Fake). Although it demonstrated less bias towards a specific class when compared to ResNet50, VGG16 did not show sufficiently consistent discriminative patterns for a robust separation between real and synthetic data, indicating instability in its classification decisions.

Despite being a lighter and more optimized architecture, MobileNetV2 also presented a challenging performance in the independent test, with an accuracy of approximately 48.5% (884 correct predictions out of 1824 samples). The confusion matrix revealed 414 True Negatives (Real) and 470 True Positives (Fake), but also 462 False Positives (real images classified as fake). This suggests an “alarmist” tendency, where the model easily suspects manipulation, even in the absence of deepfake, which can generate false alerts in practical applications.

Discussions and Analyses

The divergence between high performance in internal validation and low accuracy in independent testing is a widely documented phenomenon in the computer vision literature, known as the generalization gap (Corrêa, 2023). This discrepancy frequently occurs in deepfake detection tasks when the model fails to transfer knowledge acquired from a specific sample to data exhibiting variations in compression and noise not seen during training (Wang et al., 2020). Deep Learning models trained on specific deepfake datasets tend to suffer from over-specialization on the statistical artifacts inherent to the training sample.

The technical explanation for this stagnation of metrics in the test resides in the concept of Shortcut Learning. Convolutional Neural Networks (CNNs) have an intrinsic bias towards textures over the shape or semantics of the object (Geirhos et al., 2019). In the context of this work, it is likely that the Transfer Learning models have identified “shortcuts” in the training images, such as microscopic patterns of digital noise or specific lighting vibrations of the Celeb-DF v2 dataset, which do not represent the essence of genuine facial manipulation. This behavior suggests that the model may have focused on dataset-specific noise patterns, a vulnerability also observed by Junior (2023).

The preprocessing step, including bilinear resizing of images to 224 x 224 pixels, may have functioned as an unintentional smoothing technique. This smoothing might have eliminated synthesis artifacts that would only be noticeable at native resolutions, especially in high-quality deepfakes like those from Celeb-DF v2 (Rössler et al., 2019). The detection of synthetic manipulations is extremely sensitive to image understanding and resizing, and the fixed input resolution of classical architectures imposes a physical limit on the detection of fine details.

Another factor that contributed to the generalization gap was the structure of the training and validation data, performed by the Keras/TensorFlow function `image_dataset_from_directory`. Although this function automates data labeling and separation (80% training, 20% validation), the partition is executed at the individual file (frame) level, ignoring the temporal origin and subject identity contained in the original video. This may have resulted in “face data leakage”, where visually almost identical images were distributed between the training and validation sets, inflating the validation accuracy (88%) and masking the real inability to generalize (Lopes et al., 2026).

The computational infrastructure used, Google Colab, imposed critical limitations on RAM, disk storage, and continuous execution time. These restrictions limited the exploration of hyperparameters and the number of training epochs to only 10. Robust models like ResNet50 and VGG16 often require hundreds of epochs to fine-tune their weights on high-complexity problems (He et al., 2016). The sub-optimization of hyperparameters, due to these limitations, may have prevented the networks from breaking the memorization barrier and achieving true generalization.

The similarity of results between architectures of distinct complexities (ResNet50 and MobileNetV2) in the independent test reinforces the thesis that the obstacle lies not in the computational power of the networks, but rather in the need for data refinement (fine-tuning) and increased diversity of the training sample to break the generalization barrier. The 50% accuracy in the independent test, according to the thesis of Lopes et al. (2026), was a byproduct of identity memorization (overfitting), and not of genuine detection of synthetic artifacts, revealing that automation facilitated by `image_dataset_from_directory`, without rigorous separation by individuals, tends to mask the model’s real inability to generalize when faced with new attacks.

Ultimately, the experimental journey undertaken in this study, from class rebalancing via downsampling to the exhaustive evaluation of Transfer Learning models, reveals that detecting deepfakes in high-fidelity datasets like Celeb-DF v2 transcends simple hyperparameter optimization. The gap observed between the validation accuracy of 88% and the test accuracy of 50% should not be interpreted as a model failure, but rather as a precise technical diagnosis of the limitations of contemporary CNNs. The fallacy of accuracy in controlled environments conceals the fragility of texture artifacts that are often eliminated in preprocessing, as confirmed by the thesis of Lopes, Laroca, and Grégio (2026).

Therefore, the results presented here serve as a methodological alert, indicating that the fight against disinformation requires a paradigm shift, moving the focus from laboratory performance to robustness in real-world scenarios. The reliability of a digital security system lies not in the pursuit of high nominal metrics, but in the model’s ability to generalize synthesis patterns in a way that questions the subject’s identity. This study establishes a solid foundation for future research to incorporate uncertainty analyses and data partitions strictly based on identities, aiming to break generalization barriers and ensure that Artificial Intelligence acts as a truly integral filter in the preservation of biometric truth.

4. Conclusion

This study evaluated the effectiveness of pre-trained transfer learning models in distinguishing between real and synthetic images, specifically deepfakes. It was found that, after preprocessing and rebalancing the Celeb-DF v2 dataset, the ResNet50, MobileNetV2, and VGG16 architectures demonstrated high performance in internal validation, with ResNet50 achieving an AUC of 0.8762. However, when subjected to an independent test set, the models exhibited a significant generalization gap, with accuracies stabilizing around 50%. It was observed that this discrepancy stemmed from shortcut learning, where the models focused on digital noise patterns and specific training dataset textures, rather than genuine manipulation artifacts. Resizing the images to 224 x 224 pixels may also have smoothed or eliminated synthesis artifacts, and partitioning the data at the frame level, rather than by identity, contributed to facial data leakage, inflating validation metrics and masking true generalization capability.

The main contribution of this work lies in methodologically warning that robust deepfake detection in real-world scenarios requires more than hyperparameter optimization or simple weight transfer from generic models. The computational limitations of the development environment restricted the exploration of hyperparameters and the number of training epochs, preventing the networks from breaking the memorization barrier and achieving real generalization. For future studies, data augmentation techniques and fine-tuning of the deeper convolutional layers are suggested. The incorporation of uncertainty analyses and the use of strictly identity-based data partitions are also recommended, aiming to overcome generalization barriers and ensure that Artificial Intelligence acts as an integral filter in preserving biometric truth, effectively combating disinformation through synthetic media.

Bibliographic References

Forbes, 2023 [Referência completa não encontrada no documento original]

Junior, N.; Ediberto do J. 2023. Detecção de deepfakes utilizando redes neurais convolucionais. 46 f. Trabalho de Conclusão de Curso de Graduação em Engenharia de Computação, Campus de Sobral, Universidade Federal do Ceará, Sobral, CE, Brasil. Disponível em: <https://repositorio.ufc.br/handle/riufc/78832>. Acesso em: 31 ago. 2025.

Kaufman, D. 2022. Desmistificando a inteligência artificial. Belo Horizonte, MG, Brasil Autêntica.

MONTAÑO, R. A. N. R.: Inteligência Artificial. 1 ed. Curitiba [PR]: IESDE, 2021.

Silva, R. C. S. 2024. Prevenção contra deepfakes: desenvolvimento de um sistema de reconhecimento facial para diferenciar rostos humanos de rostos gerados por IA. Disponível em: <http://repositorio.ifam.edu.br/jspui/

Westerlund, M. 2019. The Emergence of Deepfake Technology: A Review (Volume 9, Issue 11). Disponível em <https://timreview.ca/article/1282> Acesso 15 out. 2015.

Article originating from the Final Course Project of the Specialization in Data Science and Analytics of the MBA USP/Esalq

To learn more about the course, click here and access the MBX Academy platform

Most recent

Economy, Investments And Banking

October 23, 2026

Educação Financeira, Comportamentos e Vieses na Tomada de Decisão Financeira

October 22, 2026

Challenges and Possibilities of the First School Year of Children with ASD for Teaching Work

October 22, 2026

Consumer Journey in the Decision to Purchase Infoproducts in the Digital Environment

October 22, 2026

Psychosocial Risks and Mental Health at Work in a Banking Institution

October 22, 2026

Digital Observatory of the Amazon: Data Science Applied to the Analysis of Social, Economic, Environmental, and Health Indicators

October 22, 2026

Continuous Learning as a Factor of Organizational Growth in the Perception of Collaborators

October 22, 2026

Statistical Regression in Microgrids for Operational Performance Estimation: Campusgrid Case

October 22, 2026

Forecast of Aggregate Default in the Brazilian Financial System by Sarimax Model with Macroeconomic Variables

October 22, 2026

Stakeholder Engagement Factors in International Technology Transfer to a Public Hospital

A gestão financeira pessoal representa um desafio crescente, influenciada por vieses cognitivos que afetam a tomada de decisão, mesmo em indivíduos com alto conhecimento. O presente estudo analisou a relação entre o nível de educação financeira e a ocorrência de vieses comportamentais na tomada de decisão financeira de adultos residentes no Brasil. Realizou-se um levantamento de campo de natureza aplicada, quantitativa e descritivo-correlacional, por meio da aplicação de um questionário estruturado a 163 respondentes. O instrumento contemplou variáveis sociodemográficas, conhecimento e hábitos financeiros, e situações hipotéticas para identificar vieses cognitivos. Os resultados indicaram que níveis mais elevados de educação financeira associaram-se a melhores práticas de controle financeiro, poupança e investimento, com correlações significativas com renda e escolaridade. Contudo, a presença de vieses comportamentais (ancoragem, aversão à perda, escassez, otimismo/negatividade, viés do presente, status quo, excesso de confiança e contabilidade mental) persistiu mesmo entre indivíduos com maior escolaridade e renda, sem associação estatisticamente significativa com as variáveis sociodemográficas analisadas. Concluiu-se que a educação financeira, embora essencial para a organização e o controle financeiro, não foi suficiente para neutralizar a influência de vieses cognitivos. Sugere-se a incorporação de abordagens comportamentais aos programas de educação financeira.

Palavras-chave: Comportamento financeiro; Economia comportamental; Finanças pessoais; Heurísticas cognitivas; Vieses cognitivos.

October 22, 2026

Challenges and Possibilities of the First School Year of Children with ASD for Teaching Work

The inclusion of children with Autism Spectrum Disorder (ASD) in Early Childhood Education presents significant challenges, given the difficulty of adapting to new environments and the intensity of stimuli. The objective was to analyze the teaching performance in Early Childhood Education in the initial schooling process of children with ASD. A case study was conducted, using narrative interview as a data production technique. The results revealed that the teaching performance was characterized by mediation and affection practices, with the flexibility of routine and the use of playful resources as essential strategies for the child’s development and sensory organization. The importance of the school support professional for the success of the pedagogical work was also highlighted. The child’s integral development was promoted through intentional teaching actions that prioritized environmental adaptation, strengthening of bonds, and attention to the specific needs of ASD. As a contribution, a guide of pedagogical strategies aimed at welcoming children with ASD in public daycare centers was developed.

Keywords: Early Childhood Education; School Inclusion; Teacher Narratives; Inclusive Pedagogical Practices; Autism Spectrum Disorder.

October 22, 2026

Consumer Journey in the Decision to Purchase Infoproducts in the Digital Environment

The expansion of e-commerce and technological transformation have driven the need to understand consumer behavior regarding infoproducts, whose perceived experience has become crucial for professional development. The study aimed to analyze the factors that influenced the purchase decision of infoproducts in the digital environment, considering the stages of the consumer journey. An applied, quantitative, and descriptive research was conducted through a field survey with 100 consumers from Serra Gaúcha. Data collection occurred via a structured questionnaire, using non-probabilistic sampling. The results revealed a qualified audience with strong “mobile-first” behavior. It was found that the availability of detailed information and social approval were the main attraction triggers in the pre-purchase phase. In the transactional and subsequent stages, the minimization of operational friction, the agility of immediate access, and system security proved crucial for conversion and satisfaction, although gaps in human support efficiency were identified. Retention was based on the trust generated by positive past experiences, overcoming the price factor. It was concluded that the infoproduct consumer journey operates as an integrated system, requiring alignment between technological infrastructure, content curation, and support humanization to ensure engagement.

Keywords: E-commerce; Consumer behavior; Consumer experience; Loyalty; Virtual business.

October 22, 2026

Psychosocial Risks and Mental Health at Work in a Banking Institution

Mental health in the workplace is a topic of growing relevance, impacting professional development and employee well-being. A study was conducted to identify aspects and situations that can influence absenteeism levels in a banking institution. The research involved 108 participants and used a structured questionnaire, based on the short version of COPSOQ II by Gonçalves (2021), which contained 45 closed Likert scale questions and two open-ended questions. Data were collected via an online platform and analyzed with a descriptive and quantitative approach. The results revealed a scenario that indicated the need for adjustments in the distribution of activities and in managers’ perception of behaviors and signs of disorders. Important evidence was identified in the participants’ open-ended responses. It was concluded that the study contributed to promoting the perception of a safe environment for mental health in the organization.

Keywords: Absences due to disorders; Professional development; Operational leadership; Medical leave; Mental health.

October 22, 2026

Digital Observatory of the Amazon: Data Science Applied to the Analysis of Social, Economic, Environmental, and Health Indicators

The state of Amazonas faced historical challenges in consolidating the Unified Health System (SUS), stemming from geographical barriers, socioeconomic inequalities, and structural limitations. This study described the development and implementation of a health information geoportal, designed to support monitoring, evaluation, and planning in the Amazonian context. A descriptive case study approach was adopted, based on documentary analysis and secondary data, covering the mapping of functionalities, technical infrastructure, and validation steps of the tool. The Portal was developed with open technologies and integrated multiple national databases, offering interactive dashboards, thematic maps, and a digital document repository. Among the results, highlighted were the georeferenced visualization of key indicators, the possibility of temporal comparative analyses, and the use of the platform in reviewing health plans and redirecting strategic actions. The portal can represent an advance in digital governance and evidence-based management, with potential for replication in other highly complex territories, provided it is accompanied by investments in technological infrastructure, professional training, and inter-institutional coordination.

Keywords: Amazon; Health Evaluation; Digital Health; Health Information Systems.

October 22, 2026

Continuous Learning as a Factor of Organizational Growth in the Perception of Collaborators

The impact of continuous learning and knowledge management on organizational growth was studied in a context of globalization and constant changes. The study’s objective was to analyze employees’ perception of learning culture and its importance for personal and professional development. Data were obtained through an anonymous survey, with a quantitative approach, using a form with “survey” type questions applied to 87 professionals from the technology, health, education, commerce, and services sectors. The results indicated that employees had a positive perception of the contribution of continuous learning to personal and organizational growth, recognizing its benefits for the development of new skills, innovation, and competitiveness. However, it was observed that the incentive for internal knowledge exchange and the offering of practical training still presented moderate levels of agreement. It was concluded that continuous learning significantly contributes to individual and organizational development, suggesting that companies should develop strategic policies for learning focused on new skills, aiming for greater competitive advantage and innovation.

Keywords: Continuous learning; Self-development; Organizational growth; Knowledge Management; Innovation.

October 22, 2026

Statistical Regression in Microgrids for Operational Performance Estimation: Campusgrid Case

A multiple statistical regression model was proposed to estimate the operational performance of UNICAMP’s CampusGrid microgrid, based on internal electrical quantities. The microgrid integrates photovoltaic generation, batteries, and a gas generator. Energy management system data were collected at five-minute intervals over one month. The Operational Reliability Index, defined as the ratio between supplied and demanded energy, was established as the dependent variable. The model was adjusted for the islanded operation mode, evaluating the impact of demand, state of charge, solar generation, and supplied energy. The results revealed high explanatory power, indicating that operational reliability depends on the interaction between supply and demand. It was observed that the nighttime period concentrated the highest rate of partial supply, attributed to the reduction in solar generation. It was concluded that multiple regression offers an interpretable and useful framework for microgrid monitoring and control strategy support.

Keywords: Battery storage; Photovoltaic generation; Reliability Index; Islanded operation; Energy Management System.

October 22, 2026

Forecast of Aggregate Default in the Brazilian Financial System by Sarimax Model with Macroeconomic Variables

Aggregate non-payment in the Brazilian financial system is a crucial indicator of credit risk, the prediction of which is regulated by CMN Resolutions n.° 4.557/2017 and 4.966/2021 for the prospective measurement of expected losses. Aggregate non-payment was modeled and predicted using the SARIMAX model, projecting its trajectory until December 2026 under optimistic, base, and pessimistic scenarios. 164 monthly observations (July/2012 to February/2026) and macroeconomic variables such as unemployment, Selic, IPCA, IBC-Br, and household debt were used. The specification with contemporaneous variables was selected based on information criteria and coefficient stability, and residual diagnostics confirmed statistical adequacy, with a MAPE of 3.12%. Historical non-payment ranged between 4.33% and 7.64%. The estimation identified unemployment (β = 0.2534; p < 0.001) and Selic (β = 0.2128; p < 0.001) as the main positive determinants, with strong temporal persistence (AR(1) = 0.8965). Selic showed increasing correlation with non-payment (r = 0.62 at lag 12), while unemployment demonstrated a contemporaneous effect (r = 0.33). Projections indicated non-payment between 6.30% and 7.98% in December 2026, with a range of 1.68 percentage points, highlighting direct implications for provisioning in financial institutions.

Keywords: Unemployment; Predictive modeling; Monetary policy; Loss provisioning; Time series.

October 22, 2026

Stakeholder Engagement Factors in International Technology Transfer to a Public Hospital

This study described the critical factors perceived for stakeholder engagement in international technology transfer in a public hospital, with a focus on planning. The objective was to describe these conditioning factors, according to the perception of professionals. A single case study was conducted, applying a questionnaire to 25 professionals and using content analysis, complemented by descriptive tabulations. The main obstacles identified were resistance to change (72%), funding (64%), lack of communication (60%), and bureaucracy (40%). Only 28% of respondents recognized the existence of adequate tools for stakeholder management, while 48% denied it and 24% did not know. Exploratory cross-tabulations indicated variations in the assessment of influence according to time of practice, but did not establish causal relationships. Lack of engagement was associated with delays, low adherence, waste of resources, communication failures, compromised knowledge absorption, and risks to patient safety. The conclusions were limited to the perceptions of the researched group and supported preliminary recommendations for intersectoral communication and coordination.

Keywords: Case study; Project Management; Health innovation; Stakeholders; Public sector.

Economy

October 19, 2026

End of the 6×1 scale: what do the numbers allow us to conclude?

Following the promulgation of the end of the 6×1 shift system, the debate on the shorter workday moves to implementation and requires distinguishing social benefits from economic costs. The end of the 6×1 shift system transitioned from legislative discussion to a constitutional change in Brazil. On October 16, 2026, the National Congress promulgated Constitutional Amendment No. 141/2026, originating from Proposal […]

Artificial Intelligence

October 16, 2026

Why study if AI exists?

When answers become easy, learning comes to mean more than just arriving at them. Why do I need to learn this if I can ask an artificial intelligence? This type of question has accompanied education for decades. Why learn an operation if a calculator exists? Why memorize information if it can be found […]

Tax Management

October 15, 2026

Pink tax and the gender of price

How different marketing treatments for men and women are incorporated by consumption taxation Imagine a shelf on which two razors of the same brand and intended for the same purpose are displayed side by side: one has dark packaging and promises strength, precision, and efficiency; the other comes in shades of pink or […]

Health

October 13, 2026

The missing link in health transformation

Why is innovation not enough to drive changes in organizations and health systems Digital transformation has become a strategic priority for health systems worldwide. Electronic health records, artificial intelligence, telemedicine, remote monitoring, and data analysis are now part of the strategic agenda of governments and health organizations. Paradoxically, […]

Marketing

October 09, 2026

Bitcoin or pets: what Artificial Intelligence would not see

A real-life consumer intelligence case shows what gets left out of the brief A few years ago, an American insurance company hired a specialized consumer intelligence firm to answer three questions: The insurer was also considering launching paid insurance in bitcoin. At the time, the consultancy’s workflow was as follows: the analyst created a query […]

School Management

October 06, 2026

The school is not a casino: when trading cards become currency in the playground

Why, for certain students, does school manage to offer so few experiences of achievement? The World Cup is over. Brazil left the competition, the school holidays came, the final was played and the football calendar moved on. It would be reasonable to imagine that, with the end of the tournament and the return to classes, […] would also begin

Energy Transition

September 30, 2026

Renewable energies: the Brazilian advantage

Brazil needs to invest in the capacity to transform natural resources into innovation, productivity, and development. Talking about energy is talking about development. Behind industry, transportation, services, cities, and increasingly artificial intelligence, there is an energy infrastructure that sustains the economy. Understanding a country’s energy matrix is understanding […]

Data Analysis

September 30, 2026

When the data changes, has reality changed?

An analysis of almost 20 million teaching positions-year between 2020 and 2025 shows how changes in population composition and record production can alter the interpretation of historical series Between 2020 and 2025, the number of formal teaching employment positions identified in RAIS (Annual Relation of Social Information), from the Ministry of Labor, […]

Artificial Intelligence

September 28, 2026

Artificial intelligence and the trends

Why do so many people join in on the fun with ChatGPT? I was never the trend person. I still am not. Some pass through my timeline, I see them, I find them interesting but not to the point of trying them out. This was the case, inclusive, with some recent waves of images made by artificial intelligence, like the 80s trend, the one that […]

Tax Management

September 01, 2026

Tax reform and estate planning in the countryside

Efficiency in managing taxes facilitates governance and allows for the organization of family succession. Rural wealth organization has become a strategic measure in light of the tax reform in Brazil, which came into effect this year. The planning of activities and rural wealth management, through tools such as the transition from an Individual to […]

Renata de Gaspari Valdejão Almeida

Artificial Intelligence

August 05, 2026

Inside the company and off the radar: the risks of Shadow AI

According to Microsoft, 75% of professionals already use artificial intelligence at work without company approval. The use of generative artificial intelligence is already part of the corporate routine. Tools like ChatGPT, Copilot, and other platforms help professionals produce texts, analyze data, create presentations, and accelerate tasks in practically all sectors. But this advance […]

Renata de Gaspari Valdejão Almeida

Strategy

July 02, 2026

The future as a competence

Academic interest reinforces the projection of scenarios as an essential tool for decision-making. While some companies are still trying to get the next step right, others are already working with multiple possible futures — and preparing for all of them. But after all, what does it mean to “think about the future”? Is it a trend, a strategy, or a new competence that […]

Renata de Gaspari Valdejão Almeida

Innovation

Technology

May 21, 2026

Agro SP Corridor: innovation, institutional articulation and future

Pecege commands the executive operation of the 400 km ecosystem activities that became known as the Brazilian agro Silicon Valley In 2025, in addition to taking over the management of the Piracicaba Technology Park, the Pecege Institute was also designated to command the executive operation of the State Agribusiness Innovation Corridor activities, the so-called […]

Education

April 27, 2026

How to turn a TCC into a business

New management of the Piracicaba Technology Park offers consulting to MBA graduates The new project director of the Piracicaba Technology Park (PTP), Alexandre Barreto, often says that one of the greatest rewards he has ever had in life was seeing his MBA USP/Esalq students structure companies from their final projects […]

Technology

March 20, 2026

The human repertoire in the age of AI

Critical thinking and creativity are irreplaceable in the work of a marketing professional Artificial Intelligence (AI) is capable of incredible feats, but it does not yet possess the judgment, empathy, intuition, and strategic creativity necessary to do the job of a good marketing professional. In this area, it is important to know how to use the tool to automate […]

Renata de Gaspari Valdejão Almeida

Agribusiness

Innovation

February 23, 2026

ESG as a growth strategy in a cooperative in RS

Santa Clara balances social and environmental responsibility with innovation Founded in 1911 by Italian immigrants in Carlos Barbosa, in the Serra Gaúcha region, Cooperativa Santa Clara is an example that it is possible to balance economic growth with social and environmental responsibility. From a small cheese production company, the organization has evolved into a […]

Renata de Gaspari Valdejão Almeida

Agribusiness

Innovation

January 22, 2026

Brazil free of Foot-and-Mouth Disease: what changes for Brazilian beef?

With a herd of over 200 million head of cattle, the country is currently the leader in exports. Brazil recently received, for the first time, the certification of country free of foot -and-mouth disease without vaccination, conferred by the World Organisation for Animal Health (WOAH). The seal strengthens the country’s position as a reliable supplier of meat for the […]

Renata de Gaspari Valdejão Almeida

Innovation

December 09, 2025

Innovation: the new cycle of the Piracicaba Technology Park

Instituto Pecege assumes management, betting on the creation of new businesses Starting in 2026, Instituto Pecege will assume the management of the Piracicaba Technology Park (PTP). Among its responsibilities will be the administration, operation, maintenance, and expansion of the park, in addition to fostering activities in innovation, entrepreneurship, and science and technology. The choice was made […]

Renata de Gaspari Valdejão Almeida

July 06, 2026

Large-Scale Architecture: DSR-SCRUM Change Management

Mônica Alessandra Guerios; Maria do Carmo Assis Todorov Abstract generated by ResumeAI, an artificial intelligence solution developed by Instituto Pecege focused on synthesis and writing. Architecture and urbanism are configured as essential bases in an increasingly complex and challenging urban space, where the intrinsic connection between social, environmental, and technological issues […]