Principais Coleções do Repositório
Submissões Recentes
Predição de risco de crédito com aprendizado de máquina supervisionado: um estudo de caso com dados desbalanceados
(Universidade Federal de Lavras, 2025-09-30) Melo, Rafael Almeida Pereira; Guimarães, Paulo Henrique Sales; Melo, Marcel Irving Pereira; Silva Neto, Darcy Ramos da; Pala, Luiz Otavio de Oliveira; Veloso, Manoel Vitor de Souza
This work evaluates the application of penalized Logistic Regression via LASSO in credit risk
prediction, focusing on the treatment of imbalanced data, a common characteristic in financial
datasets where the proportion of defaulters is usually low. A total of 36 synthetic datasets were
generated through simulation, with different combinations of sample size and proportion of defaulters. The model was calibrated using cross-validation and dynamic adjustment of the cut-off
(decision threshold), without employing resampling techniques such as SMOTE (Synthetic Minority Over-sampling Technique, which generates synthetic examples of the minority class) or
undersampling (reducing the majority class to balance the distribution). Instead, resampling
procedures with replacement from the original dataset were adopted to generate different scenarios of sample size and imbalance. The results showed robust performance in metrics such as
AUC, F1 Score, and balanced accuracy, even in scenarios with severe imbalance, that is, when
the proportion of defaulters is extremely low (between 1% and 5%). The LASSO penalization
contributed to the automatic selection of relevant variables, while maintaining model interpretability. As a practical contribution, the study demonstrates that, with careful adjustments, it
is possible to obtain statistical models that are simple, effective, and compliant with regulatory
requirements, providing reliable support for credit risk management. Furthermore, the findings
can be generalized to other contexts involving imbalanced data, expanding the applicability of
the proposed approach.
Respostas morfofisiológicas ao déficit hídrico em gramíneas com metabolismos C3, C4 e intermediário
(Universidade Federal de Lavras, 2025-05-12) Botelho, Alexandre dos Santos; Barbosa, João Paulo Rodrigues Alves Delfino; Livramento, Darlan Einstein do; Alvarenga, Joyce Pereira; Lara, Márcio André Stefanelli
Grasses (Poaceae) are the foundation of human and animal food systems and play a strategic role in natural ecosystems and large-scale biomass production. In the context of climate change, water limitation has become one of the main constraints to plant productivity, highlighting the need to better understand the mechanisms that confer resilience to stress. This dissertation aimed to investigate how different photosynthetic types — C₃, C₄, and intermediate C₃–C₄ — influence the morphophysiological and productive responses of tropical grasses under drought conditions. The study was conducted under controlled conditions and included the evaluation of leaf water status, cellular integrity, morphometric traits and productivity in four species: Triticum aestivum (C₃), Urochloa brizantha (C₄), Homolepis isocalycia, and H. longispicula (C₃–C₄ intermediates). Results showed that photosynthetic metabolism shapes distinct drought response strategies: while the C₄ species exhibited high productivity, its sharp decline under stress reflected a strong dependence on water availability; the C₃ species showed structural limitations and increased leaf senescence; and the intermediate species exhibited contrasting behaviors, with H. longispicula notably maintaining active leaf production under stress. The main finding of this study is the evidence that traits associated with functional stability — such as the preservation of vegetative architecture and continuous growth under water restriction — may already be present in C₂ grasses. These traits reinforce the role of intermediate species as functional transitions between the C₃ and C₄ extremes, contributing to the understanding of photosynthetic evolution in Poaceae. As a future perspective, further investigation is recommended on other Homolepis species and on additional ecophysiological variables, including water use efficiency, radiation use efficiency, leaf area index, and relative growth rates.
Avaliação do potencial antimicrobiano e anti-inflamatório de filmes bioativos à base de pectina
(Universidade Federal de Lavras, 2025-08-29) Silva, Monique Suela; Piccoli, Roberta Hilsdorf; Batista, Karla de Aleluia; Oliveira, Juliano Elvis de; Gonçalves, Michelle Carlota; Castro, Andreísa Teixeira de; Oliveira, Juliano Elvis de; Pereira, Alcilene de Abreu
The growing bacterial resistance to antibiotics and the limitations of conventional treatments in wound healing have driven the search for alternative and sustainable therapies. In this context, bioactive dressings based on pectin incorporated with essential oils (EOs) have shown great potential. The skin, a structurally complex organ with vital functions, is vulnerable to injuries that compromise its integrity, making wound healing a multifaceted process influenced by factors such as infection, age, and nutritional status. Chronic wounds, often associated with comorbidities like diabetes and aging, have a significant clinical and economic impact, demanding effective solutions. Modern dressings, particularly bioactive ones, interact with the wound environment to promote healing. Pectin, a natural polysaccharide with biocompatible properties, forms stable gels, absorbs exudate, and serves as a matrix for cell adhesion. Although lacking inherent antimicrobial activity, pectin can be functionalized with bioactive compounds such as EOs to enhance its therapeutic efficacy. Essential oils of rosemary, citronella, lavender, and tea tree exhibit antimicrobial, anti-inflammatory, and antioxidant properties through multitarget mechanisms, reducing the emergence of resistant strains. Their actions include disrupting bacterial membranes, inhibiting inflammatory mediators, and protecting against oxidative stress. When integrated into dressings, these compounds contribute to tissue regeneration, provide analgesic effects, and support the management of topical infections. Nanotechnology, particularly through nanoemulsions, improves the stability, skin permeability, and bioavailability of EOs, overcoming challenges like volatility and poor water solubility. These formulations are suitable for pharmaceutical, food, and cosmetic applications. This work is grounded in the principles of Integrative and Complementary Health Practices (PICS), as established by Brazil's National Policy, aligning with the goal of promoting natural, accessible, and holistic healthcare. Thus, the development of sustainable pectin-based films incorporating EOs represents an effective, low-cost therapeutic alternative with positive implications for Brazil’s Unified Health System (SUS), fostering innovation in wound care and supporting the scientific advancement of aromatherapy.
Técnicas de resfriamento de arroz cozido alteram a formação de amido resistente
(Universidade Federal de Lavras, 2025-08-25) Carvalho, Rafaella Ribeiro Sâmia de; Bastos, Sabrina Carvalho; Botelho, Flávia Barbosa Silva; Pinheiro, Elisângela Elena Nunes
Rice (Oryza sativa L.) is a cereal of global importance, whose nutritional value is significantly affected by processing and cooking. RS, a non-digestible starch fraction with beneficial effects on metabolic and intestinal health, is mainly formed during the retrogradation of cooked and cooled starch and is influenced by intrinsic rice factors such as amylose concentration, fiber, lipids, and proteins, as well as extrinsic factors like cultivation conditions and added components during preparation, such as oils. This study investigated the modulation of resistant starch (RS) in rice, considering the influence of different genotypes, the type of oil used in preparation, and cooling methods. The methodology involved the characterization of rice genotypes (BRSEsmeralda, BRSMG Caçula, CMG1590, and commercial varieties) regarding amylose content, proximate composition of raw and prepared grains, as well as the quantification of phenolic compounds and antioxidant capacity. The rice preparation technique was standardized, altering the oil used for sautéing (avocado oil or soybean oil) and the cooling method (room temperature, refrigerator at 4°C for 24h, and freezer at -15°C for 24h). RS quantification was performed by a standardized enzymatic method (Association of Official Analytical - AOAC, 2002). It wasfound that amylose content is important, but the grain matrix, including fiber content, apparently modulates RS formation, sometimes overriding high amylose content. Post-cooking cooling resulted in significant increases in RS for most genotypes, except black rice, which showed a reduction (up to -22.36% with soybean oil in the freezer). The type of oil (avocado vs. soybean) showed a marginal impact on RS formation; however, avocado oil, with its superior nutritional profile, can add value to the meal without compromising RS. It is concluded that grain composition and cooling are determining factors in RS formation, offering practical strategies to optimize health benefits.
Análise da base dos preços de milho para as principais praças de negociação em Minas Gerais por modelos de séries temporais
(Universidade Federal de Lavras, 2025-08-29) Tavares, Victória Barbosa; Oliveira, Izabela Regina Cardoso de; Simão, Sérgio Domingos; Rodrigues, Letícia Lima Milani; Ávila, Ednilson Sebastião de; Guimarães, Paulo Henrique Sales
Brazil has become one of the leading producers and exporters of corn, which is one of the most important economic activities in the national agricultural sector. Corn is a cereal with high nutritional value, widely used as a raw material in various industry segments, such as the food sector, agricultural production, and even in the manufacture of renewable fuels, such as ethanol. Understanding how the price of this commodity behaves is essential for trading this commodity, as it enables the maximization of revenues and the reduction of risks for both buyers and sellers. In the derivatives market, the basis refers to the difference between the price of the commodity in the physical market and its price in the futures market. The weakening or strengthening of this basis may indicate scenarios that are more or less favorable to the trajectory of the product’s prices. The weakening or strengthening of this basis may indicate scenarios that are more or less favorable to the product’s price trajectory. Given this, this study aims to adjust time series models to analyze the volatility of the corn price basis in the state of Minas Gerais, comparing this variation in the different physical markets in this region. The Autoregressive (AR) model and the Generalized Autoregressive Conditional Heteroscedasticity (GARCH) model were used. The data used corresponds to the daily prices of corn in physical markets in different locations in Minas Gerais, as well as futures market prices obtained from B3, covering the period from July 16, 2018, to December 31, 2024. The adjusted models performed well, allowing predictions to be made for a period of 7 and 15 days in each market analyzed. The modeling showed that the corn base in Minas Gerais has persistent volatility, with shocks extending over several periods, allowing for better prediction of variations and supporting more secure marketing decisions. Based on the evaluation criteria applied, it was concluded that the models were adequate, adjusting well to the daily base data for corn in the markets of the state of Minas Gerais during the period studied.