A comparative study of ensemble machine learning algorithms for brucellosis disease prediction

Abstract

Brucellosis, caused by Brucella spp., is a global public health concern, particularly in underdeveloped regions. Cattle, predominantly infected with B. abortus, encounter reproductive challenges, reduced productivity, and fertility issues. Effective control measures, including serological tests like iELISA (indirect Enzyme-linked Immunosorbent Assay) are vital. This research harnesses machine learning techniques, encompassing AdaBoostM1, Vote, Bagging, and LogitBoost, to forecast Brucella infection in cattle, utilizing comprehensive data sourced from Qazvin, Iran. Detailed model descriptions are provided, highlighting AdaBoostM1 as the optimal choice, boasting a robust 75% correlation, low RMSE (Route Mean Square Error), MAE (Mean Absolute Error), and a commendable Kappa Statistics score of 0.4965. Ensemble machine learning demonstrates significant potential in Brucellosis prediction, adept at handling intricate datasets, and enhancing predictive accuracy. AdaBoostM1 stands out as the preferred model, offering valuable insights for Brucellosis prediction and contributing to the enhancement of disease control strategies.