Machine learning to detect behavioural anomalies in dairy cows under subacute ruminal acidosis

Abstract

et al.. Machine learning to detect behavioural anomalies in dairy cows under subacute ruminal acidosis. A B S T R A C TSickness behaviour is characterised by a lethargic state during which the animal reduces its activity, sleeps more and at times when normally awake, reduces its feed and water intake, and interacts less with its environment. Subtle modifications in behaviour can materialise just before clinical signs of a disease. Recent sensor developments enable continuous monitoring of animal behaviour, but the shift to abnormal animal activity remains difficult to detect. We explored the use of Machine Learning (ML) to detect abnormal behaviour from continuous monitoring. We submitted 14 cows (Bos taurus) to Sub-Acute Ruminal Acidosis (SARA), a disease known to induce changes in behaviour. Another 14 control cows were not submitted to SARA. We used a ruminal bolus to monitor pH and detect when a cow experienced SARA. We used a positioning system to infer an animal's activity based on its position in relation to specific elements in the barn (feeder, resting area, and alleys). We tested several ML algorithms: K Nearest Neighbours for Regression (KNNR); Decision Tree for Regression (DTR); MultiLayer Perceptron (MLP); Long Short-Term Memory (LSTM); and an algorithm where activity is assumed to be similar from one day to the next. First, we developed ML models to predict activity on a given day from the previous 24 h, considering all cows together. Then, we calculated the error between observed and predicted values for a given cow. Finally, we compared the error to a threshold chosen to optimise the distinction between normal and abnormal values. KNNR performed best, detecting 83% of SARA cases (true-positives), but it also produced 66% of false-positives, which limits its use in practice. In conclusion, ML can help detect anomalies in behaviour. Further improvements could probably be obtained by applying ML on very large datasets at animal rather than group level.

Review: Application and Prospective Discussion of Machine Learning for the Management of Dairy Farms

“…Attempts to detect subacute ruminal acidosis by analyzing behavioral anomalies with ML algorithms have failed [ 57 ]. Although it was possible to detect 83% of cases using k -nearest neighbor regression, the results were not useful in practice due to a false positive rate of 66% [ 57 ].…”

“…Particularly, rumination and feeding behavior are possible indicators for improving dairy cow management. Wagner et al [ 57 ] evaluated whether ML algorithms could be useful to predict subacute ruminal acidosis from positioning data that reflect cows’ activity. They reported that, among the tested ML algorithms, k -nearest neighbor performed best, with 83% true positives; unfortunately, the false positive alert rate was 66%.…”

Combination of Sensor Data and Health Monitoring for Early Detection of Subclinical Ketosis in Dairy Cows

“…Machine Learning approaches have been heavily used in animal behavioural/disease assessment over the last few years. Take as an example [29], where the authors applied different methods for detection of subacute ruminal acidosis in dairy cows. In their results, k-Nearest Neighbors showed the best results, outperforming deep learning methods and decision trees.…”

Changes in social and feeding behaviors, activity, and salivary serum amyloid A in cows with subclinical mastitis

“…Advances in image analysis now allow automated recognition of individuals within a herd (Andrew et al, 2020), and accurate identification of health-related abnormal behaviors, such as foot disease (Gu et al, 2017). The identification of behaviors associated with subclinical disease (before the development of, or without, clinical symptoms) may therefore find application in future diagnostic software algorithms targeted at early disease monitoring in dairy cows (e.g., Wagner et al, 2020).…”