Anomaly Detection in Domestic Animals using Real-Time Veterinary Data

Appleton, Charlotte and Jewell, Christopher and Rowlingson, Barry (2025) Anomaly Detection in Domestic Animals using Real-Time Veterinary Data. Masters thesis, Lancaster University.

Text (2025appletonmphil)
2025appletonmphil.pdf - Published Version
Download (18MB)

Abstract

This thesis delves into anomaly detection in domestic animals using SAVSNet veterinary consultation data. The main aim is to create an automatic surveillance system that takes the consultation from recorded, through a method of analysis that is effective for the data and displaying the results back to the veterinary surgeons and stakeholders. This thesis explores Gaussian Processes as well as mixed models to decide on an anomaly detection methodology before creating an automatic surveillance system and using Shiny Apps and Tableau to display the results. The research reveals that using a mixed model to model Spatio-temporal data was more effective than applying just a Gaussian Process as we were able to remove consistent characteristics including seasonality from the data and model departures from these characteristics using a Gaussian Process. One of the key findings is that using the mixed model was successful in detecting outbreak style patterns as it led to the confirmation of a Canine Enteric Coronavirus outbreak in January 2020. These insights contribute to the veterinary surgeons and stakeholders ways of working in that they have direct access to current rates for different Main Presenting Complaints within their county and Nationwide. Having this information accessible aids in the preparation for the veterinary surgeons e.g. having medicines readily available for seasonal periods. The implications of this research are significant for domestic animal research as it leaves a tool that can be monitored for anomalies and thus can help influence decisions regarding domestic animal health. It is also an offering to the general public and owners of the animals as the dashboard is also created with those with lesser statistical knowledge in mind, alongside an emphasis on people with accessibility issues. In summary, this thesis effectively applies anomaly detection methodologies and through the creation of an automated surveillance system is able to relay results to veterinary surgeons, stakeholders, other researchers and the general public to improve research knowledge in this area.