Comparison of machine learning methods to predict udder health status based on somatic cell counts in dairy cows

Tania Bobbo,Cristian Taccioli,Mauro Penasa,Stefano Biffani,Martino Cassandro

doi:10.1038/s41598-021-93056-4

Tania Bobbo, Cristian Taccioli + Show 3 more

Open Access

https://doi.org/10.1038/s41598-021-93056-4

Copy DOI

Abstract

Bovine mastitis is one of the most important economic and health issues in dairy farms. Data collection during routine recording procedures and access to large datasets have shed the light on the possibility to use trained machine learning algorithms to predict the udder health status of cows. In this study, we compared eight different machine learning methods (Linear Discriminant Analysis, Generalized Linear Model with logit link function, Naïve Bayes, Classification and Regression Trees, k-Nearest Neighbors, Support Vector Machines, Random Forest and Neural Network) to predict udder health status of cows based on somatic cell counts. Prediction accuracies of all methods were above 75%. According to different metrics, Neural Network, Random Forest and linear methods had the best performance in predicting udder health classes at a given test-day (healthy or mastitic according to somatic cell count below or above a predefined threshold of 200,000 cells/mL) based on the cow’s milk traits recorded at previous test-day. Our findings suggest machine learning algorithms as a promising tool to improve decision making for farmers. Machine learning analysis would improve the surveillance methods and help farmers to identify in advance those cows that would possibly have high somatic cell count in the subsequent test-day.

Highlights

Bovine mastitis is one of the most important economic and health issues in dairy farms
The aim of the present study was to compare the performance of ML algorithms in predicting the udder health status of cows at TD n + 1 using cow and milk information collected at the previous TD n
Prediction models developed within eight different ML methods, namely Linear Discriminant Analysis (LDA), Generalized Linear Model (GLM) with logit link function, Naïve Bayes (NB), Classification and Regression Trees (CART), k-Nearest Neighbors, Support Vector Machines (SVM), Random Forest (RF) and Neural Network (NN) were trained and tested on 80% of the data (14,755 records) to identify the best udder health prediction method based on data previously recorded on cow and milk

Summary

Introduction

Bovine mastitis is one of the most important economic and health issues in dairy farms. Thanks to the recent implementation of novel milk-testing technologies based on flow cytometry in different laboratories of the Italian Breeders Associations (Rome, Italy), differential somatic cell count (DSCC), i.e., the ratio of neutrophils and lymphocytes to total S CC13, is monthly recorded with the aim of improving the identification of the mammary gland status of dairy cows. Routine data collection during monthly recording procedures and access to large datasets including information on herd, cows and milk composition suggest the possibility to use machine learning (ML) classification algorithms to predict the udder health status of cows. Prediction of subclinical mastitis using information (i.e., animal information, milk production and composition) recorded on the previous test-day (TD) in the frame of the monthly recording procedure across several dairy herds have not been investigated yet Such analysis would improve the surveillance methods and help farmers to identify in advance those cows that would possibly present high SCC level in the subsequent TD. The aim of the present study was to compare the performance of ML algorithms in predicting the udder health status of cows (healthy or mastitic according to SCC below or above a predefined threshold) at TD n + 1 using cow and milk information collected at the previous TD n

Objectives

Methods

Results

Conclusion