Application of Machine Learning for Predicting Brucellosis Disease in Dairy Cattle

Abstract

Abstract Brucellosis, a zoonotic disease caused by Brucella bacteria, has emerged as a significant concern in Bangladesh, affecting both animals and humans, with economic repercussions and health risks. This paper addresses the urgent need for effective disease management in a country where agriculture and livestock are pivotal to the economy. The study focuses on identifying risk factors for brucellosis in dairy cattle and their correlation with factors collected from Central Cattle Breeding and Dairy Farm (CCBDF) and Military Dairy Farms (MDF) in Bangladesh. Serum and milk samples, along with demographic data, were collected and analyzed. To address the challenge of imbalanced data, the study introduces the application of the "SMOTE" Weka filter for the first time in Bangladesh's veterinary profession. This balances the dataset, enhancing the accuracy of predictive modeling. Machine learning models, specifically Multilayer Perceptron (MLP) and J48 decision tree algorithms, are employed for brucellosis prediction. MLP achieved a Correct Classification (CC) rate of 95.0801%, while J48 achieved 94.0503%, demonstrating their efficacy. Sensitivity analysis identifies key attributes, with retained placenta being the most significant, aiding in prioritizing control measures. Association Rules reveal patterns and relationships among disease-related factors, enhancing understanding and proactive prevention.