DeepBrucel: A Deep Learning Approach for Automated Risk Detection of Brucellosis in Cattle Farms in Ecuador

Abstract

An automated risk model for Brucellosis detection in cattle farms, termed DeepBrucel, was developed and validated. A comprehensive survey encompassing 51 variables related to farm characteristics, management practices, and reproductive pathologies was administered across 632 cattle farms in Ecuador. The extensive dataset thus obtained was utilized to implement and compare classifiers based on regression, neural networks, and deep learning methodologies. A wide-ranging primary experimentation protocol enabled the identification of critical variables and the optimal topology for the neural networks. Superior performance was exhibited by a deep neural network model with three hidden layers, which achieved an impressive accuracy of 98.4% in predicting Brucellosis risk. DeepBrucel, now publicly available, provides a highly accessible and robust tool for the diagnosis and control of Brucellosis in cattle farms.