Work

Machine Learning-Based Prediction of Brucella Seropositivity and Associated Factors among Rwandan Dairy Cattle with Reproductive Disorders

Year: 2026
Type: preprint
Abstract: Brucellosis poses significant economic and zoonotic challenges to Rwanda’s dairy sector. This study assessed factors associated with Brucella seropositivity among high-risk dairy cattle and benchmarked machine learning (ML) algorithms for predictive surveillance. A cross-sectional survey was conducted from June to December 2025 across four Rwandan districts (Nyagatare, Kayonza, Bugesera, and Rubavu). Serum samples and survey data were collected from 274 purposively selected dairy cattle with documented reproductive disorders from 75 farms. Serum samples were screened using the Rose Bengal test (RBT), and RBT-positive samples were subsequently tested using an indirect ELISA for confirmation. Multivariable logistic regression was used to assess factors associated with seropositivity (p < 0.05), and five ML models (Random Forest, Logistic Regression, KNN, Decision Tree, SVM) were evaluated using 5 metrics: Precision, recall, F1 score, accuracy, and Area under the Curve (AUC). Overall seropositivity rate among cattle with reproductive disorders was 10.2% (28/274), with spatial clustering in Nyagatare (19.8%) and Kayonza (7.3%). Multivariate analysis identified district, farm position, farmer education, disease awareness, and vaccination as key independent predictors. Among ML algorithms, Random Forest performed best, achieving 89.09% accuracy, 89.74% precision, 94.59% recall, 92.11% F1-score, and an AUC of 86.19%. Non-linear models consistently ranked district, education, and farm position as top predictive features. Among the selected dairy cattle with reproductive disorders, Brucella seropositivity was associated socio-demographic factors, whereas vaccination and disease awareness exert strong protective effects. Random Forest demonstrated promising predictive performance for identifying seropositive cattle within the study population. The findings from this study provide field-based evidence to inform risk-based brucellosis surveillance in Rwanda and other countries with comparable epidemiological and dairy production settings.
Language: English
Field: Medicine
Open Access status: green
OpenAlex ID: W7225912099
Cites: 21
Cited by:
Related to: 10