Sensor Data Classification for the Indication of Lameness in Sheep

Al-Rubaye, Zainab, Al-Sherbaz, Ali, McCormick, Wanda and Turner, Scott (2018) Sensor Data Classification for the Indication of Lameness in Sheep. In: 13th International Conference, CollaborateCom 2017, Edinburgh, UK.

Text Accepted Version Available under the following license: Creative Commons Attribution Non-commercial No Derivatives. Download (826kB)

Abstract

Lameness is a vital welfare issue in most sheep farming countries, including the UK. The pre-detection at the farm level could prevent the disease from becoming chronic. The development of wearable sensor technologies enables the idea of remotely monitoring the changes in animal movements which relate to lameness. In this study, 3D-acceleration, 3D-orientation, and 3D-linear acceleration sensor data were recorded at ten samples per second via the sensor attached to sheep neck collar. This research aimed at determining the best accuracy among various supervised machine learning techniques which can predict the early signs of lameness while the sheep are walking on a flat field. The most influencing predictors for lameness indication were also addressed here. The experimental results revealed that the Decision Tree classifier has the highest accuracy of 75.46%, and the orientation sensor data (angles) around the neck are the strongest predictors to differentiate among severely lame, mildly lame and sound classes of sheep.

Item Type:	Conference or Workshop Item (Paper)
Keywords:	sensor data classification, machine learning, decision tree, lameness detection, sheep
Faculty:	Faculty of Science & Engineering
Depositing User:	Ian Walker
Date Deposited:	03 Oct 2018 14:54
Last Modified:	01 Apr 2022 14:40
URI:	https://arro.anglia.ac.uk/id/eprint/703641

Actions (login required)

Edit Item