Special Issue: Engineering Advances in Precision Livestock FarmingResearch PaperFeed-forward and generalised regression neural networks in modelling feeding behaviour of pigs in the grow-finish phase☆,☆☆
Introduction
Feeding behaviour of grow-finish pigs can be used to inform producers of both health status and stress level. Many parameters have been studied to better understand feeding behaviour of pigs, including feed intake, meal length, meal interval, number of meals, and total time spent eating (Morgan et al., 2000, Nienaber et al., 1990, Nienaber et al., 1991, Quiniou et al., 2000). Most of these measurements have been obtained from feeding systems that allow only one pig to feed at any given time, which is not representative of commercial production where pigs typically feed in a group setting (Brown-Brandl, Rohrer, & Eigenberg, 2013).
Feeding behaviour is dependent on several environmental and genetic factors, including but not limited to temperature, humidity, gender, breed, and time of day. Deviations from normal feeding behaviour may indicate that grow-finish pigs are experiencing a stressful event, such as illness, issues with feed quality, or heat-related stress. Models of feeding behaviour could be used as a management tool to assess stress levels within a population and to identify sick animals.
Several different approaches have been used to analyse and model feeding behaviour of pigs. Linear regression and analysis of variance models have been used extensively (Brown-Brandl et al., 2013, Nienaber et al., 1990, Nienaber et al., 1991, Quiniou et al., 2001). However, application of these methods is limited due to complex, non-linear relationships between multiple input variables (Comrie, 1997). Gaussian models (Morgan et al., 2000), three-process random models (Berdoy, 1993), and logistic models (Tolkamp & Kyriazakis, 1999) have also been applied to predict feeding behaviour. There are two major drawbacks to these types of models. They tend to be very complex, and they require prior knowledge of relationships between input variables, i.e. a predefined functional form for the model.
Artificial neural networks (ANN) have emerged as a powerful tool in applications where complexity of relationships between inputs and outputs makes formulating a comprehensive mathematical model nearly impossible (Hecht-Nielsen, 1989). An ANN is a set of computing systems that imitates learning abilities of neurons in the brain. Artificial neural network models have the ability to handle large amounts of noisy data, without requiring prior information on model form. An additional advantage of ANN models over other statistical methods is that they require less training data (Paola & Schowengerdt, 1995). The ability to learn by example makes a neural network a very flexible and powerful tool.
This study focused on application of ANN models for prediction of feeding behaviour patterns of pigs during the grow-finish phase. Abilities of feed-forward neural networks (FFNN) and generalised regression neural networks (GRNN) to predict feeding behaviour of grow-finish pigs throughout the year, using time of day and temperature humidity index (THI) as inputs, were compared.
Section snippets
Feed-forward neural network (FFNN)
Feed-forward neural networks are one of the most popular ANN models used in engineering applications. The network architecture and learning algorithm of a FFNN can be viewed as a generalisation of the well-known least-mean-square (LMS) algorithm (Haykin, 2007). Figure 1 shows the architecture of a typical FFNN, which includes three layers: an input layer, a hidden layer which is responsible for performing intermediate computations, and an output layer. The input signal propagates through the
Neural network training and parameter optimisation
Temperature humidity index and time of day (time period) were used as inputs in development of several ANN models to predict feeding behaviour of grow-finish pigs. Environmental temperatures are known to affect feeding behaviour of pigs (Quiniou et al., 2000). Ideally, barn temperatures would have been used in the development of our ANN models. However, barn temperature data were only available for part of this study. Thermal conditions inside the barn were approximated by THI due to its strong
Conclusions
Artificial neural network models have become increasingly popular in many different fields due their ability to elucidate complex, non-linear relationships between parameters. The focus of this study was to identify an accurate and efficient neural network model for predicting feeding behaviour of grow-finish pigs based on time of day and THI. Four different ANN models were trained using electronic feeder data from 1923 grow-finish pigs, and their performance was assessed using data from four
Funding
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.
Acknowledgements
The authors would like to thank K. Simmerman for excellent technical assistance and the U.S. Meat Animal Research Center swine crew for outstanding husbandry.
References (29)
Defining bouts of behaviour: A three-process model
Animal Behaviour
(1993)- et al.
Analysis of feeding behaviour of group housed growing-finishing pigs
Computers and Electronics in Agriculture
(2013) - et al.
Multilayer feedforward network with nonpolynomial activation function can approximate any function
Neural Networks
(1993) - et al.
A model for monitoring the condition of young pigs by their drinking behaviour
Computers and Electronics in Agriculture
(2005) - et al.
Early detection of health and welfare compromises through automated detection of behavioural changes in pigs
The Veterinary Journal
(2016) A scaled conjugate gradient algorithm for fast supervised learning
Neural Network
(1993)- et al.
Analysis of the feeding behavior of pigs using different models
Physiology & Behavior
(2000) A new fruit fly optimization algorithm: Taking the financial distress model as an example
Knowledge-based Systems
(2012)- et al.
Voluntary feed intake and feeding behaviour of group-housed growing pigs are affected by ambient temperature and body weight
Livestock Production Science
(2000) - et al.
Introduction to multi-layer feed-forward neural networks
Chemometrics and Intelligent Laboratory Systems
(1997)
To split behaviour into bouts, log-transform the intervals
Animal Behaviour
Artificial neural networks
Development of a livestock feeding behaviour monitoring system
Transactions of the American Society of Agricultural Engineers
Cited by (0)
- ☆
Mention of a trade name, proprietary product, or specified equipment does not constitute a guarantee or warranty by the USDA and does not imply approval to the exclusion of other products that may be suitable.
- ☆☆
The USDA is an equal opportunity provider and employer.