Determinants of student performance in ENEM in São Paulo

Paulo, Determinantes do Desempenho de Estudantes no Enem em São

doi:10.22167/2675-6528-202603290

Article

October 13, 2026

Determinants of student performance in ENEM in São Paulo

Determinants of Student Performance in the Enem in São Paulo

Muriel Figueiredo Maia; José Erasmo Silva

DOI: 10.22167/2675-6528-202603290

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 study analyzed factors associated with student performance in the National High School Exam (Enem), considering the historical relevance of educational assessment in Brazil. Its objective was to evaluate the predictive performance of the XGBoost model and identify the variables most associated with the participants’ average performance. Microdata from Enem from 2015 to 2023 were used, restricted to the state of São Paulo, organized into student, school, and socioeconomic questionnaire levels. After cleaning and standardization, a final database with 1,886,735 records was constructed. The XGBoost model was configured with hyperparameters to balance predictive capacity and generalization, including native handling of categorical variables. The results indicated that XGBoost showed performance compatible with educational studies. Interpretation by internal contribution values revealed the predominance of institutional and socioeconomic factors in explaining performance, with emphasis on the school’s administrative dependence and variables related to the family context. The research demonstrated that predictive approaches based on trees were capable of capturing relevant patterns without the need for traditional hierarchical models, offering a methodological alternative for understanding educational inequalities. It was concluded that the combination of large data volume, machine learning techniques, and interpretation by contributions allowed the identification of relevant factors for Enem performance and provided subsidies for reflections on educational policies.

Keywords: Machine Learning; Educational performance; Enem; Socioeconomic factors; XGBoost.

1. Introduction

The evaluation process has an extensive trajectory, with records of practices initiated in China aimed at selecting officials, and over time it has come to assume different functions, including measuring student performance. This process was consolidated with studies carried out in England and France and, with the development of influential models in the United States throughout the 20th century (Carvalho, 2017). In Latin America, research on school performance advanced more recently, between the 1980s and 1990s, when Brazil began to develop instruments aimed not only at verifying student performance but also at analyzing the factors that influence it. In this context, evaluations such as the National System for the Evaluation of Basic Education [SAEB] emerged and, as one of its branches, the National High School Examination [Enem] (Carvalho, 2017).

Created in 1998, Enem’s initial objective was to assess students at the end of basic education, and it was later reformulated in 2009 to also serve as a mechanism for accessing higher education (INEP, 2025). School performance, in turn, has been widely studied, and one of the most recognized models is Walberg’s (1984), which identified nine factors determining student achievement, organized into three dimensions: individual characteristics, school factors, and environmental factors.

The influence of this model can be observed in later studies, such as the analysis of socioeconomic and demographic factors through ANOVA, highlighting the impact of gender, race, family income, school type, region, and parents’ education on performance (Carvalho, 2017). Additional studies, such as those conducted in the Northeast region, identified elements that favor or hinder performance, including parents’ education, access to technological resources, age, and income, through linear regression (Souza et al., 2021). In addition to these works, there are also studies that used multilevel models to analyze the performance of third-year high school students, demonstrating the importance of considering the hierarchical structure of the data (Andrade et al., 2007).

Despite the methodological diversity, a gap is observed: a large part of these studies focuses on traditional explanatory models, while predictive approaches capable of dealing with non-linear relationships and complex interactions are still less explored in the analysis of performance in the Enem. Furthermore, although the data present a hierarchical structure involving students and schools, the objective of this study is not to estimate random effects, but to identify general patterns and variables most associated with average performance.

In this scenario, it becomes important to understand which factors stand out in explaining the average performance of ENEM participants and how a tree-based predictive model, such as XGBoost, can capture these patterns even without resorting to traditional hierarchical models. The relevance of this approach lies in the fact that, although multilevel models are widely used in educational research, predictive methods capable of handling non-linear relationships and complex interactions are still less explored in this type of analysis. The use of XGBoost offers a methodological alternative that allows for flexible exploration of the data structure, contributing to the understanding of factors associated with performance and providing evidence that can support discussions on educational policies and strategies aimed at reducing inequalities. Thus, the study aims to evaluate the predictive performance of XGBoost in identifying the variables most associated with students’ average performance, simultaneously considering individual and institutional information present in the exam’s microdata.

2. Material and Methods

The study was conducted in the Visual Studio Code (VS Code) integrated development environment, using the Python programming language. Libraries such as pandas and NumPy were used for data manipulation and processing, scikit-learn for splitting into training and testing sets, cross-validation, and metric calculation, and XGBoost for building the predictive model. The data used originated from the microdata of the National High School Exam (Enem), made available by the National Institute for Educational Studies and Research Anísio Teixeira (INEP/MEC), containing information about the student, school, test location, results, and socioeconomic profile of the participants.

The data collection covered the period from 2015 to 2023, restricted to participants linked to schools in the state of São Paulo. The sample considered high school students who had not completed this stage in the reference year or who completed it in the year of the exam application, aiming to standardize the analysis group. The database was organized in hierarchical layers, covering information about the student, the school, the test, and the socioeconomic questionnaire.

Cleaning procedures were applied, including the exclusion of absent participants in all tests, the removal of records with zero scores in all areas, and the elimination of rows with null values in the column referring to the municipality’s name. The socioeconomic questionnaire variables, originally coded by letters, were converted to numerical values. Similarly, the gender variable was standardized, converting “F” and “M” to 0 and 1.

These conversions had an operational purpose, adapting the categorical variables to the model’s execution environment without altering their intrinsic relationships. The target variable, named MEDIA_GERAL, was calculated as the arithmetic mean of the scores obtained by the participants in the five evaluated areas of the exam. After preprocessing, the final database consisted of 1,886,735 records and 39 variables, with the general average being the only numerical one.

A statistical exploration of the target variable was performed, observing an approximately continuous and symmetric distribution, with no need for additional transformations for model training. Multicollinearity among socioeconomic variables was assessed by calculating the Variance Inflation Factor (VIF), with values below the threshold of 5, the criterion adopted to ensure estimation stability (O’Brien, 2007).

Moderate correlations between socioeconomic variables were observed, especially those related to the possession of durable goods and residential structure, as well as between parents’ schooling. These correlations indicated the need to interpret the importance of the variables in the XGBoost model jointly, and not in isolation.

The XGBoost model was chosen, a tree-based approach, for its ability to handle non-linear relationships and for not depending on low collinearity between predictors. XGBoost, an efficient implementation of “boosting”, sequentially builds decision trees, correcting errors from previous ones and incorporating regularization mechanisms (Bruce et al., 2025).

The XGBoost model was configured to balance predictive performance and generalization. A reduced learning rate was used to mitigate overfitting, and the maximum tree depth was limited. Sampling techniques for observations and variables were employed to increase model robustness and stability. Native handling of categorical variables was enabled, and settings were adopted to ensure reproducibility (XGBoost Developers, 2024).

For comparison purposes, a Linear Regression model was developed, trained and evaluated with the same dataset (70% train, 30% test) and random seed as the XGBoost. Categorical variables were converted into ordinal numerical codes for the Linear Regression, which represented a methodological simplification for this comparison (Souza et al., 2021).

Additionally, a naive baseline model, which predicted the mean of the target variable observed in the training set, was used as a reference to quantify the predictive gain of XGBoost compared to the absence of any predictive model. This approach allowed for the contextualization of the performance of more complex models.

The stability of the results was evaluated through k-fold cross-validation, with five partitions (k = 5), applied to the XGBoost model, maintaining the original hyperparameters. This procedure ensured the reliability of the estimates, minimizing the dependence on a single data split and reinforcing the robustness of the evaluation.

The importance of the variables was estimated using SHAP (SHapley Additive exPlanations) values, a technique that quantifies the contribution of each variable to the model’s prediction. These values were calculated internally by XGBoost, allowing for the measurement of the average impact of variables on the prediction, including those categorical variables treated natively, ensuring consistency with the trained model.

3. Results and Discussion

The evaluation of the predictive performance of the XGBoost model revealed a coefficient of determination (R²) of approximately 0.32, indicating that the model was able to explain about 31% of the observed variability in student performance. In practical terms, this means that approximately one-third of the differences in performance among participants can be attributed to the variables included in the model, such as socioeconomic profile, school characteristics, and demographic data. The remaining two-thirds of the variability stem from factors not captured by the dataset, such as pedagogical and emotional aspects, and the individual preparation of each student. Although this R² does not represent an extremely high level of explanation, it is compatible with the results of initial models in the field of education, where performance is influenced by a multiplicity of factors that are not always directly observable in the available data.

The error metrics provided a more detailed understanding of the model’s accuracy. The mean squared error (MSE) was approximately 6.539, and the root mean squared error (RMSE) was about 80.86 points. Although MSE, due to its quadratic scale, does not have a direct interpretation, RMSE, by returning the error to the original scale of the grade, facilitates understanding. An RMSE of 80.86 points suggests that the grade predicted by the model deviates from the student’s actual grade, on average, by approximately 80 points, more or less. For example, a student with an actual grade of 550 points would tend to have a prediction between 470 and 630 points.

The Mean Absolute Error (MAE) was approximately 59.29 points, measuring the average distance between the predicted score and the actual score without squaring the differences. The distinction between MAE and RMSE is relevant: MAE better reflects the error most students face, while RMSE, more sensitive to large errors, tends to be higher when there are instances of significant error. In this study, the fact that the MAE (59.29) is noticeably lower than the RMSE (80.86) indicates that, although most predictions have a relatively small margin of error, there is a smaller group of students for whom the model errs by considerably larger margins, which elevates the RMSE.

The comparison of XGBoost’s performance with other models revealed its superiority. Compared to the naive baseline model, which predicts the overall average for all students, XGBoost’s RMSE dropped from 99.05 to 82.35 points, a reduction of 16.9%. This improvement demonstrates that the inclusion of socioeconomic, school, and demographic information consistently enhances prediction. Compared to Linear Regression, XGBoost also achieved a superior fit, with an R² of 0.3088 versus 0.2514 and an RMSE of 82.35 versus 85.70 points. This result was expected, as XGBoost is capable of capturing non-linear relationships and interactions between variables that a linear model cannot represent.

Nevertheless, the moderate difference between XGBoost and Linear Regression suggests that a significant portion of the relationship between explanatory variables and performance is approximately linear, being captured even by a simpler approach. The Pearson correlation between the actual values and those predicted by XGBoost was 0.56, indicating a moderate to strong association. This value, at a favorable intermediate level, reinforces that the model consistently follows the observed variation in grades, even with a moderate R², which is important for understanding general performance patterns.

To verify the stability of the results and ensure that the performance was not an artifact of a specific data split, a k-fold cross-validation was performed, with five partitions. The average R² obtained in the five rounds was 0.3080, with a standard deviation of 0.0018, and the average RMSE was 82.33 points, with a standard deviation of 0.23. This minimal variation between rounds, with R² ranging between 0.3062 and 0.3098 and RMSE between 82.10 and 82.56 points, indicates that the model’s performance remained stable regardless of how the data was split. This consistency reinforces the reliability of the evaluation performed and the robustness of the XGBoost model for this dataset.

The interpretation of the importance of the variables was performed using SHAP (SHapley Additive exPlanations) values, a technique that estimates the average contribution of each variable to the prediction of a student’s grade relative to the overall average grade. The higher the SHAP value of a variable, the greater its influence on the model’s predictions. The values were obtained through XGBoost’s internal calculation of contributions, allowing the average impact of each variable to be measured even with categorical variables handled natively by the algorithm, which ensures consistency between the model and the interpretation of the results.

The analysis of the importance of the variables revealed a predominance of institutional and socioeconomic factors. The variable TP_DEPENDENCIA_ADM_ESC, which represents the administrative dependence of the school, showed the greatest influence on the model, with an average impact of 23.23 points on the predicted score. This means that, on average, the administrative dependence of a student’s school alters the score predicted by the model by about 23 points, more or less, an effect considerably larger than that of any other variable analyzed. This finding reinforces the relevance of school management as a determining factor in student performance.

The distribution of the overall grade by the school’s administrative dependency corroborates this observation, showing that federal and private schools present higher performance medians than state and municipal schools. Although the performance dispersions between the groups are similar, the difference in medians is consistent with the magnitude of the impact identified by SHAP for this variable. This pattern suggests that the administrative structure of the educational institution plays a crucial role in student performance, possibly reflecting differences in resources, infrastructure, and pedagogical approaches.

Among other notable factors, socioeconomic variables from the student questionnaire appeared, such as Q006 (family income), Q024 (computer ownership), Q001 (father’s schooling) and Q002 (mother’s schooling), and Q013 (freezer ownership). The contributions of these variables ranged from 4.56 to 9.31 points, indicating that family context and living conditions exert a significant influence on the results. Parents’ schooling (Q001 and Q002) and family income (Q006) are classic indicators of cultural and economic capital that can influence the study environment and access to complementary educational resources, corroborating existing literature on educational inequalities.

The possession of durable goods, such as computers (Q024) and freezers (Q013), also proved relevant, reflecting the family’s socioeconomic level and access to resources that can facilitate study and development. The TP_ESCOLA variable, which indicates the type of school (public, private, or abroad), also showed a relevant impact, with an average of 3.96 points, reinforcing the importance of the institutional environment. Individual attributes, such as age group (6.09 points), race or color (3.59 points), sex (1.61 points), and nationality, although contributing to the explanation of the phenomenon, showed a lower relative weight compared to institutional and socioeconomic factors.

The comparison of the results of this study with previous research demonstrates important convergences on the factors associated with educational performance. The study by Souza et al. (2021), conducted in the Northeast region, identified significant differences between socioeconomic groups, with a gap of 88 points between low-income and non-low-income students. This pattern directly dialogues with the findings of the present work, in which family income (Q006) appears among the variables with the greatest contribution in the model. Furthermore, Souza et al. (2021) identified that parents’ schooling, access to computer and internet, and school type influence performance, factors that also emerge among the most important variables of XGBoost, such as TP_DEPENDENCIA_ADM_ESC, Q001, Q002, Q024, and TP_ESCOLA.

It is important to note that Souza et al. (2021) used Linear Regression as the analysis method, and although XGBoost showed superior predictive performance to Linear Regression in this work, the difference was moderate. This suggests that a considerable part of the relationship between the explanatory variables and student performance is already captured by a simpler linear approach. The fact that similar factors emerge as relevant in both the linear modeling and the non-linear approach adopted in this work reinforces the robustness of these findings, rather than evidencing a pronounced methodological advantage of XGBoost in terms of isolated predictive accuracy.

The study by Carvalho (2017), which analyzed performance in the Enem writing section, also highlighted socioeconomic and demographic factors, including family income, guardians’ schooling, school type, race, and region. Although Carvalho’s (2017) focus was on a specific area of the test, the variables that proved relevant are equivalent to those that appeared among the main contributions of the predictive model used here. This reinforces that socioeconomic and institutional inequalities manifest consistently across different components of the Enem, indicating a transversal influence of these factors on students’ overall performance.

Finally, research based on multilevel modeling by Andrade and Laros (2007) showed that 17% of the variance in school performance can be attributed to the school level, demonstrating that institutional differences explain a relevant part of the results. This conclusion converges with the present study, in which the most influential institutional variable was precisely the administrative dependence of the school (TP_DEPENDENCIA_ADM_ESC). The XGBoost’s ability to capture nonlinear relationships and complex interactions between variables, without the need for traditional hierarchical models, offers a valuable methodological alternative for understanding educational inequalities, even with the limitation of not estimating random effects.

Despite the robustness of the XGBoost model, some methodological limitations must be considered. The comparison with Linear Regression required the conversion of categorical variables into ordinal codes, a simplification that may have underestimated the performance of this model. Furthermore, the absence of more detailed pedagogical, emotional, and contextual information in the database limited the predictive reach of XGBoost, whose R² of approximately 0.31 reflects this gap. However, the value of XGBoost in this work lies less in producing drastically better predictions than traditional methods and more in offering an approach capable of capturing non-linear relationships without compromising the consistent identification of the institutional and socioeconomic factors that effectively explain educational performance.

In summary, the research results indicated that factors related to school structure and family context had the greatest weight in predicting student performance in the Enem, while individual characteristics, such as sex, race, and age, showed less influence. The XGBoost model demonstrated superior predictive performance to the naive baseline model and Linear Regression, and k-fold cross-validation confirmed the stability and reliability of these findings. The identification of the school’s administrative dependence and socioeconomic variables such as family income and parental education as the main factors associated with students’ average performance meets the study’s objective and provides important subsidies for reflections on educational policies aimed at reducing inequalities.

4. Conclusion

This study sought to evaluate the predictive performance of the XGBoost model in identifying the variables most associated with the average performance of students in the National High School Exam, considering individual and institutional information. It was found that the XGBoost model presented predictive performance compatible with the educational area, explaining about 31% of the observed variability in grades and outperforming reference models such as linear regression and the naive baseline model. The stability and reliability of these findings were confirmed through k-fold cross-validation. The analysis of variable importance, performed using internal contribution values, revealed the predominance of institutional and socioeconomic factors. The administrative dependence of the school stood out as the variable with the greatest influence, indicating that school management plays a crucial role in performance. Variables such as family income, parents’ education level, and ownership of durable goods also proved relevant, reinforcing the influence of family context and living conditions. Individual attributes, such as age group and race or color, had a lower relative weight. These results converge with existing literature, which points to the consistency of socioeconomic and institutional inequalities in educational performance.

Despite the robustness of the XGBoost model, some limitations were identified. The comparison with linear regression required the conversion of categorical variables into ordinal codes, which may have underestimated the performance of this model. Furthermore, the absence of more detailed pedagogical, emotional, and contextual information in the database limited the predictive reach of XGBoost. However, the main contribution of the study lies in demonstrating that predictive approaches based on trees are capable of capturing relevant patterns and consistently identifying the institutional and socioeconomic factors that explain educational performance, even without the need for traditional hierarchical models. For future studies, it is suggested to include pedagogical, emotional, and contextual variables, as well as to apply more robust coding techniques for categorical variables in comparative models, aiming to expand the predictive capacity and comparability between approaches. This work thus offers a valuable methodological alternative and subsidies for the improvement of educational policies aimed at reducing inequalities.

Bibliographic References

Andrade, J.M.; Laros, J.A. 2007. Fatores associados ao desempenho escolar: estudo multinível com dados do SAEB/2001. Psicologia: Teoria e Pesquisa 23(1): 33-42.

Bruce, P.; Bruce, A.; Gedeck, P. 2025. Estatística prática para cientistas de dados: 50+ conceitos essenciais usando R e Python. 2ed. Alta Books, Rio de Janeiro, RJ, Brasil.

Carvalho, F.E.D. 2017. Fatores socioeconômicos associados ao desempenho dos estudantes na prova de redação do Exame Nacional do Ensino Médio (Enem). 179f. Tese (Doutorado em Educação) – Universidade Federal do Ceará, Faculdade de Educação, Programa de Pós-Graduação em Educação, Fortaleza, CE, Brasil. Disponível em: https://repositorio.ufc.br. Acesso em: 20 set. 2025.

Instituto Nacional de Estudos e Pesquisas Educacionais Anísio Teixeira [INEP]. 2025. Enem – apresentação. Disponível em: https://www.gov.br/inep/pt-br/areas-de-atuacao/avaliacao-e-exames-educacionais/enem. Acesso em: 13 set. 2025.

O’BRIEN, R.M. 2007. A Caution Regarding Rules of Thumb for Variance Inflation Factors. Quality & Quantity 41(5): 673–690.

Souza, A.E.; Santos, L.M.S.; Larruscaim, I.M.; Besarria, C.N. 2021. Determinantes do desempenho no Enem na região Nordeste: uma análise de dados em painel do período de 2015 a 2019. Revista Brasileira de Estudos Regionais e Urbanos 15(4): 690–711. DOI: 10.54766/rberu.v15i4.915.

Walberg, H.J. 1984. Improving the productivity of America’s schools. Educational Leadership 41: 19-27.

XGBoost Developers. 2024. XGBoost Documentation – Parameters. Disponível em: https://xgboost.readthedocs.io/en/stable/parameter.html. Acesso em: 11 jan. 2026.

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

October 13, 2026

“Brand”: economic feasibility for LGBTQIA+ sustainable fashion in São Paulo

October 13, 2026

Predictors of death in the elderly with severe acute respiratory syndrome confirmed by Influenza A(H1N1)pdm09

October 13, 2026

Building beauty from an early age: makeup and skincare consumption behaviors in Generation Alpha

October 13, 2026

A comparative overview of electric vehicle fleets in Brazil and the European Union

October 13, 2026

Population and Immigration in Japan: Projections with Time Series and Demographic Causal Analysis until 2044

October 13, 2026

Digital marketing and engagement on YouTube: strategies for real crime channels in Brazil

October 13, 2026

Use of Unsupervised Machine Learning techniques for planning remuneration review work for retirement

October 13, 2026

Sales Management for the Performance of a Real Estate Company

October 13, 2026

Delineo: Adaptation of a diffusion model for sketch-guided user interface prototyping

The current fashion landscape, marked by excessive “fast fashion” consumption and the exclusion of dissident bodies, has highlighted the need for brands focused on awareness, “upcycling”, and textile transmutation. In this context, the study aimed to analyze the economic feasibility of a sustainable and inclusive fashion brand, targeted at the LGBTQIA+ audience in the city of São Paulo. The research was characterized as a quantitative case study, based on the application of a questionnaire to measure consumption behavior and the projection of financial scenarios for a five-year period. The results revealed a viable average ticket of R$200.00 and confirmed the operation’s profitability with a Net Present Value of R$47,699.49, an Internal Rate of Return of 343.70%, and a Modified Internal Rate of Return of 115.80%. Sensitivity analysis proved the project’s robustness, demonstrating positive returns even in pessimistic simulations. The venture proved viable, indicating that the union of sustainability values, social economy, and an efficient financial structure results in a low-risk business.

Keywords: Investment analysis; Circular economy; Entrepreneurship; Socio-environmental impact; Apparel.

October 13, 2026

Predictors of death in the elderly with severe acute respiratory syndrome confirmed by Influenza A(H1N1)pdm09

Infection by the influenza A(H1N1)pdm09 virus can progress to Severe Acute Respiratory Syndrome (SARS), with a higher risk of death in the elderly. The objective was to analyze predictive factors of death in cases of SARS confirmed for influenza A(H1N1)pdm09, among the elderly aged 60 years or older, considering individual and contextual factors of the Administrative Regions (RA) of the State of São Paulo (ESP), and to compare the performance of regression models. A descriptive and analytical study was conducted in ESP, from 2023 to 2024, using data from SIVEP-Gripe and the SEADE Foundation. 714 cases confirmed by RT-PCR for influenza A(H1N1)pdm09 in residents of ESP aged 60 years or older were included. Sociodemographic, clinical, hospitalization, treatment, vaccination, and residence location variables were analyzed. Binary logistic regression and hierarchical logistic regression were used, with death as the outcome. Of the 714 cases, 202 resulted in death. The final binary logistic regression model associated a higher chance of death with the male sex, advanced age, oxygen saturation below 95%, ICU admission, and use of invasive ventilatory support. The presence of cough and the use of antiviral were associated with a lower chance of death. The hierarchical logistic model showed similar performance, with a small residual contextual contribution from the RAs. It was concluded that the binary logistic regression model was sufficient and simpler to represent the data, being adopted as the final model.

Keywords: Influenza A(H1N1)pdm09; Hierarchical logistic modeling; Death from Influenza A(H1N1)pdm09; Logistic Regression; Severe Acute Respiratory Syndrome.

October 13, 2026

Building beauty from an early age: makeup and skincare consumption behaviors in Generation Alpha

The female interest in aesthetics manifests in various societies, being amplified by the advancement of the internet and social media, which have driven the emergence of digital influencers. The study analyzed how digital influencer marketing impacts the consumption behavior and aesthetic perception of Generation Alpha girls (3 to 15 years old) in the state of São Paulo, from the perspective of their parents and/or guardians, focusing on the early consumption of makeup and skincare and the construction of aesthetic standards. To this end, a bibliographic research was conducted, a quantitative online and in-person survey was applied, and an interview was held with a dermatologist. The results indicated that interest in beauty arises early, with encouragement from guardians. The young girls showed a desire for belonging, and makeup items presented a slight predominance over skincare items. The majority of participants believed that influencer marketing impacts purchasing decisions. It was concluded that interest in beauty and encouragement of self-care are naturalized from an early age, and that influencer marketing contributes to the formation of consumption desires, although the non-probabilistic sample limits generalizations.

Keywords: Adolescents; Children; Aesthetics; Influencers; Social Media.

October 13, 2026

A comparative overview of electric vehicle fleets in Brazil and the European Union

Road transport contributes to approximately 20% of global CO2 emissions, driving the search for cleaner mobility solutions, such as electric vehicles. The objective of this study was to compare the electric vehicle markets in Brazil and the European Union, contrasting their policies, consumer adoption, and infrastructure development. A comparative descriptive research design was employed, analyzing quantitative data from the Brazilian Electric Vehicle Association, National Association of Automotive Vehicle Manufacturers, European Alternative Fuels Observatory, and European Environment Agency for the period from 2015 to 2024, complemented by qualitative analysis of policy documents. The results revealed a fundamental structural divergence: Brazil showed exponential growth of 80.7% per year (2022-2025), with hybrids comprising approximately 66% of sales and driven by tax incentives that leverage the ethanol bioeconomy. The European Union, in turn, showed expansion in a consolidation phase at an annual rate of 8.35% (2008-2025), with battery electric vehicles exceeding 40% of new registrations, driven by binding regulatory mandates. Infrastructure density analysis indicated that the European Union’s public charging network (55.3 points per thousand alternative fuel vehicles) was double that of Brazil’s (27.9 points per thousand electrified vehicles). Environmental quantification confirmed that Brazil’s predominantly renewable energy matrix (64 gCO2/kWh) resulted in substantially lower well-to-wheel emissions compared to the European Union’s matrix (246 gCO2/kWh). It was concluded that the Brazilian model, with technology-neutral tax incentives and leveraging biofuel infrastructure, can accelerate hybrid adoption in emerging economies with low public spending. For mature markets, the European Union’s model, with regulatory mandates, has established a dominant battery electric vehicle market, albeit with a higher fiscal cost.

Keywords: Renewable energy; Market; Qualitative; Quantitative.

October 13, 2026

Population and Immigration in Japan: Projections with Time Series and Demographic Causal Analysis until 2044

Population aging and low fertility rates present complex demographic challenges in Japan, accompanied by the expansion of foreign immigration as a response to labor supply constraints. The study aimed to analyze Japanese demographic trends and project their evolution until 2044, with an emphasis on the relationships between population decline, fertility, and immigration. The ARIMA and Prophet time series models were applied to historical data from the World Bank, the National Institute of Population and Social Security Research of Japan, and the Immigration Services Agency of Japan, covering the period from 2003 to 2023 for the population series and from 2014 to 2024 for the immigration series. The analysis included descriptive statistics, Pearson correlation, augmented Dickey-Fuller stationarity test, and Granger causality tests, in addition to predictive evaluation using MAE and MAPE metrics. The results showed the continuation of population reduction and the growth of the immigrant contingent in projections up to 2044. Prophet presented lower MAE and MAPE values on the test sets, although the short duration of the immigration series limited generalization capacity. Temporal causality tests did not identify statistically significant evidence of precedence between population and immigration. It was concluded that immigration can contribute to mitigating labor supply constraints, but it does not, in isolation, constitute a sufficient mechanism to reverse Japan’s population decline.

Keywords: Population decline; Demography; Fertility; Migration flows; Time series.

October 13, 2026

Digital marketing and engagement on YouTube: strategies for real crime channels in Brazil

The phenomenon of true crime has consolidated itself as a growth niche on Brazilian YouTube, driven by the expansion of digital media and public fascination. However, many creators have faced difficulties in achieving sustainable growth. This study analyzed how digital marketing strategies influenced the growth, engagement, and monetization of Brazilian true crime channels on the YouTube platform. A mixed-methods research approach was adopted, combining qualitative and quantitative methods, with an exploratory and descriptive nature. Four Brazilian channels in the niche were analyzed through documentary analysis of content, collection of public metrics via Social Blade and Noxinfluencer, and the application of questionnaires to four creators and 104 viewers. The results showed that the highest-performing channels adopted distinct strategies, such as a linear and contextualized narrative approach, personalized emotional communication, a technical-informative stance, or provocative opinion-based content. Convergence was observed between creators and the audience regarding the centrality of storytelling for retention, with most viewers stating that the way the story was told determined continued viewing. Creators indicated average retention time as the most predictive metric for sustainable growth, although 100% depended exclusively on AdSense for monetization, revealing financial vulnerability. It was concluded that success in the niche depends on the integration of digital marketing, consistent narrative identity, and ethical responsibility.

Keywords: creator economy; digital engagement; content marketing; storytelling; true crime.

October 13, 2026

Use of Unsupervised Machine Learning techniques for planning remuneration review work for retirement

The sizing of personnel and the management of retirement flows in federal public bodies constituted complex strategic challenges. The study aimed to apply unsupervised Machine Learning techniques, specifically Clustering (K-means) and Multiple Correspondence Analysis (MCA), to improve the planning of functional reviews in a Brazilian federal autarchy. The methodology consisted of processing a validated database of 487 active servers. Initially, quantitative variables, such as age, contribution time, and remuneration, were subjected to the K-means algorithm, resulting in the formation of five clusters, supported by the Elbow Method and Silhouette Coefficient. Subsequently, the groups were integrated with qualitative variables via MCA, projected onto a perceptual map that explained 74.81% of the data’s inertia. The results revealed that clusters 0 and 3 represented profiles of greatest pension criticality, concentrating all beneficiaries of tenure bonus and the majority of extinct positions. The research generated a phased Action Plan, prioritizing the 75 servers in these clusters who met the retirement requirements by 2025, with a similar guideline for 2026. In contrast, cluster 2 presented lower risk, with inactivity predicted only from 2031 onwards. The model proved to be a widely favorable tool for auditing and remediation, although its full effectiveness depends on more rigorous data governance in public service.

Keywords: Clustering; Multiple Correspondence Analysis (MCA); Retirement; Clusterization.

October 13, 2026

Sales Management for the Performance of a Real Estate Company

Sales management plays a strategic role in organizational performance, especially in real estate development sectors, characterized by commercial complexity. This study aimed to analyze the contribution of sales management practices to the commercial and organizational performance of a company in the real estate sector, considering its operating model and commercial structure. The research was characterized as exploratory, with a qualitative and quantitative approach, adopting a single case study. Data were collected through triangulation, using documentary research and a structured questionnaire applied to the outsourced sales team. The results indicated that the company structured its commercial management practices with monthly and annual goals, marketing campaigns, and incentive systems. Performance was monitored by quantitative indicators, such as sales volume and goal achievement percentage. In the analyzed period, 179 sales were made, corresponding to 74.6% of the established annual goal. Opportunities for improvement were observed in the alignment between commercial planning, performance monitoring, and strategy effectiveness. The brokers’ perception pointed to limitations in the incentive systems, such as the absence of progressive commissions and the duration of bonuses. It was concluded that, although sales management influences organizational performance, there are weaknesses in the alignment between effort, reward, and sales force motivation. Strengthening systematic monitoring, improving communication, defining more realistic goals, and implementing progressive incentives were recommended to enhance commercial results.

Keywords: Sales strategies; Sales force; Commercial management; Performance indicators; Real estate sector.

October 13, 2026

Delineo: Adaptation of a diffusion model for sketch-guided user interface prototyping

The evolution of artificial intelligence and deep neural networks has revolutionized human-machine interactions. Generative models, while promising for image synthesis, face challenges in prototyping user interfaces (UI) by failing to respect spatial structure. To address this divergence, a conditioned latent diffusion approach was proposed, integrating a control architecture with a large-scale model fine-tuned for mobile interfaces, which supported UI sketches to guide image inference. The training dataset comprised approximately 12,400 UI images and sketches from public sources, pre-processed and enriched with synthetic data and textual instructions. The approach utilized a Stable Diffusion 3.5 Large model as a base, adapted with ControlNet canny edges and a LoRA adapter. The approach’s feasibility was validated by quantitative evaluation, employing a normalized semantic similarity (NSS) metric based on Sentence-BERT, and by qualitative analysis with 20 industry experts. The results confirmed that the method surpassed the original model’s capabilities. However, the evaluation highlighted that the effectiveness in logical interface structuring still faced aesthetic degradation and loss of functional elements. It was concluded that, although the technology proves its technical feasibility for optimizing the transition between requirements and code implementation, the automation of creative processes demands architectural improvements to achieve the visual rigor required by the software development industry.

Keywords: Low-rank adaptation; Machine learning; Conditioned latent diffusion; Deep neural networks.

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 […]

Project Management

September 25, 2026

AI accelerates answers; who makes the decisions?

The speed of technology only becomes an advantage when it is accompanied by the capacity to understand, verify, and assume responsibility for the results. Imagine a team that starts producing in an afternoon the analyses that previously required a week. The change seems to justify the investment in artificial intelligence. But, before counting the gain, it is worth asking who examined […]

Innovation

September 14, 2026

We do not just read words

As new codes, technologies, and forms of communication expand what we need to learn to interpret. Being literate is usually understood initially as: recognizing letters, forming words, comprehending sentences, and producing texts. For a long time, it was mainly from this domain that we defined what it meant to know how to read and, consequently, to participate in different spaces of society. […]

Social Networks

September 14, 2026

The effect of influence on consumption

What the Youpix/Nielsen 2026 survey shows about the advancement of the Creator Economy In 2025, 80% of Brazilian consumers bought some product recommended by influencers. In 2026, the rate reached 81%. The stability at such a high level indicates that purchasing by influence is no longer a behavior restricted to certain audiences. The […]

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 […]