# bTB-diagnostics

This project trains a Histogram Boosted Regression Tree model on data from Bovine Tuberculosis (bTB) testing and cattle herd metadata to predict the risk of bTB outbreak.

This can be used to improve the herd-level sensitivity or specificity of the diagnostic test and also to analyse the risk factors involved in predicting bTB outbreaks.

The project consists of a number of Jupyter Notebooks:
(i) Data_Curation* -- processes the various inpiut data into a matrix for model training.
(ii) bTB-Diagnostic_2020_v4_crossVal+tuning* -- code that trains the various models.
(iii) bTB-Diagnostic_2020_final_model* -- code that performs various analysis on the models.
(iv) Vet_Data_Analysis -- code that performs some extra analysis on the veterinary data.

Further details can be found in the preprint (paper in sumbission for peer review) at: https://arxiv.org/abs/2404.03678