Kauane Ferreira da Silva

Kauane Ferreira da Silva

Auto-generated profile, based on publicly available information

Kauane Ferreira da Silva

Universidade Federal do Mato Grosso

Research Focus & Metrics

Academic Career Length

1Year (2026-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
10 August 2026 in Revista de Informática Teórica e Aplicada
Verified
Journal Article
Open access
PDF file
LALeonardo Jardim Antunes
Leonardo Jardim Antunes
AMAna Flávia Da Guia Miranda
Ana Flávia Da Guia Miranda
JJJoão Batista Gonçalves Costa Júnior
João Batista Gonçalves Costa Júnior
CACaroline Lima Amorim
Caroline Lima Amorim
MCMaria Eduarda Garcia Corrêa
Maria Eduarda Garcia Corrêa
JAJuliana Fonseca Antunes
Juliana Fonseca Antunes
Body weight estimation is an important activity in sheep management; however, direct weighing using scales may be limited by cost and operational constraints, making it necessary to seek feasible alternatives for this process. In this context, this article presents one of the first systematic comparisons of Machine Learning algorithms for predicting the body weight of Santa Inês Ewes based on morphometric measurements collected manually under field conditions. Data from 80 females evaluated in Campo Verde, Mato Grosso, Brazil, were analyzed. A formal variable selection pipeline was conducted by combining Pearson correlation analysis and the Variance Inflation Factor (VIF), which revealed severe multicollinearity among all morphometric variables (VIF ranging from 148 to 1,434), supporting the selection of thoracic circumference as the primary predictor (r = 0.92). Modeling was conducted in Python, and five regression models — Linear Regression, Random Forest, XGBoost, SVR, and MLP — were evaluated using K-Fold cross-validation (k = 10), with hyperparameter tuning performed through GridSearchCV. Results are reported as mean ± standard deviation across folds. Linear Regression achieved the best performance (R2 = 77.44% ± 13.47%, MAE = 3.021 kg), outperforming SVR, MLP, Random Forest, and XGBoost, suggesting a predominance of linear relationships in the sample. As a practical contribution, a simple estimation equation was derived: Weight(kg) = 1.298T horacicCircum f erence(cm) − 64.547, enabling producers to estimate body weight using only a measuring tape, without the need for conventional scales.

1 January 2026 in Ciência Animal Brasileira
Verified
Journal Article
Open access
KSKauane Ferreira da Silva
Kauane Ferreira da Silva
PSPedro Lucas Soldatelli
Pedro Lucas Soldatelli
RFReinaldo Santana Ferreira e Ferreira
Reinaldo Santana Ferreira e Ferreira
GFGuilherme Emanuel Ferreira Ferraz
Guilherme Emanuel Ferreira Ferraz
DADéborah Galvão Peixôto Guedes de Araújo
Déborah Galvão Peixôto Guedes de Araújo
JJJoão Batista Gonçalves Costa Junior
João Batista Gonçalves Costa Junior
Resumo A Inseminação Artificial em Tempo Fixo (IATF) é a principal biotecnologia utilizada nos sistemas de bovinocultura de corte no Brasil (93,3 % das inseminações em 2021), pois sincroniza a ovulação, supera o desafio da detecção de estro e melhora a eficiência reprodutiva. Diante da diversidade de protocolos hormonais disponíveis e das diferentes respostas reprodutivas, este estudo avaliou o desempenho preditivo de diferentes protocolos de IATF em fêmeas Nelore de rebanhos comerciais no estado de Mato Grosso, Brasil, por meio da comparação entre modelagem estatística clássica e aprendizado de máquina. Um conjunto de dados composto por 28.691 registros da estação de monta de 2023 foi analisado. O diagnóstico de prenhez (0/1) foi modelado utilizando regressão logística e modelos de aprendizado de máquina (Redes Neurais do tipo Perceptron Multicamadas). A regressão logística identificou o protocolo, o escore de condição corporal (ECC), a categoria animal e a aplicação de GnRH como preditores significativos (p < 0,05), sendo o ECC o fator com maior associação à probabilidade de prenhez. Em contraste, os modelos de aprendizado de máquina apresentaram desempenho preditivo moderado, com o Modelo Integrado superando a média dos modelos individuais (ROC-AUC = 0,721) e demonstrando maior estabilidade em comparação aos modelos específicos por protocolo (ROC-AUC = 0,694). Enquanto a regressão logística forneceu estimativas de efeito interpretáveis e suporte à inferência biológica, o aprendizado de máquina aumentou a acurácia preditiva ao capturar relações não lineares e interações complexas. Esses resultados indicam uma abordagem sinérgica entre modelos estatísticos clássicos e de inteligência artificial como estratégia para otimizar programas de IATF e aumentar a eficiência reprodutiva em sistemas comerciais de bovinos de corte.

1 January 2026 in Ciência Animal Brasileira
Verified
Journal Article
Open access
KSKauane Ferreira da Silva
Kauane Ferreira da Silva
PSPedro Lucas Soldatelli
Pedro Lucas Soldatelli
RFReinaldo Santana Ferreira e Ferreira
Reinaldo Santana Ferreira e Ferreira
GFGuilherme Emanuel Ferreira Ferraz
Guilherme Emanuel Ferreira Ferraz
DADéborah Galvão Peixôto Guedes de Araújo
Déborah Galvão Peixôto Guedes de Araújo
JJJoão Batista Gonçalves Costa Junior
João Batista Gonçalves Costa Junior
Fixed-time artificial insemination (FTAI) is the primary reproductive biotechnology used in Brazilian beef cattle systems, accounting for 93.3 % of inseminations in 2021, as it synchronizes ovulation, addresses the challenge of estrus detection, and improves reproductive efficiency. Given the diversity of hormonal protocols available and their varying reproductive performance, this study evaluated the predictive performance of different FTAI protocols in Nelore females from commercial herds in Mato Grosso, Brazil, by comparing classical statistical modeling with machine learning approaches. A dataset comprising 28,691 records from the 2023 breeding season was analyzed. Pregnancy outcome (0/1) was modeled using logistic regression and machine learning models based on multilayer perceptron neural networks. Logistic regression identified protocol, body condition score (BCS), animal category, and GnRH administration as significant predictors (p < 0.05), with BCS showing the strongest association with pregnancy probability. In contrast, the machine learning models showed strong predictive performance, with the integrated model outperforming the average performance of the individual models (ROC-AUC = 0.721) and demonstrating greater stability than the protocol-specific models (ROC-AUC = 0.694). While logistic regression provided interpretable effect estimates and supported biological inference, machine learning improved predictive accuracy by capturing nonlinear relationships and complex interactions. These findings suggest that a synergistic approach combining classical statistical models and artificial intelligence may optimize FTAI programs and enhance reproductive efficiency in commercial beef cattle systems.