|CROSS, AMANDA - South Dakota State University|
|CASSADY, JOSEPH - South Dakota State University|
Submitted to: World Congress of Genetics Applied in Livestock Production
Publication Type: Proceedings
Publication Acceptance Date: 10/16/2017
Publication Date: 2/11/2018
Citation: Keel, B.N., Cross, A.J., Brown-Brandl, T.M., Cassady, J.P., Rohrer, G.A. 2018. A feed-forward neural network for modeling feeding behavior of group-housed grow-finish pigs with respect to thermal conditions. In proceedings: World Congress of Genetics Applied in Livestock Production, 11-16 Feb 2018, Auckland, New Zealand. Volume Species-Porcine 1, 240. Available: https://www.wcgalp.org/proceedings/2018
Technical Abstract: Feeding patterns of pigs in the grow-finish phase have been investigated for use in management decisions, identifying sick animals, and determining genetic differences within a herd. Development of models to predict swine feeding behavior has been limited due the large number of potential environmental factors involved and complex relationships between them. Artificial neural networks have been proven to be an effective tool for mapping complicated, nonlinear relationships between inputs and outputs. However, they have not been applied to swine feeding behavior prediction. In this study, we developed a feed-forward neural network model to predict feeding behavior of group-housed pigs in the grow-finish phase throughout the year, using time of day and temperature humidity index as inputs. The fruit fly optimization algorithm was applied to automatically select optimal parameters for the network. Feeding behavior of pigs in a grow-finish facility was captured using an electronic monitoring system, where feeders were equipped with antennas that pinged radio-frequency identification (RFID) tags of pigs at the feeder every twenty seconds. The model was calibrated on data from 1,923 pigs in the grow-finish facility from 2008 to 2014. After calibration, predictive ability of the model was tested using data from four additional grow-finish groups reared in the same facility from 2014 to 2016. In three of the four validation groups, the model exhibited strong predictive ability, with correlations between predicted and observed feeding behaviors ranging from 0.626 to 0.742. Large deviations between predicted and observed behaviors in the fourth validation group were likely the result of an outbreak of pneumonia, which demonstrates the potential for the model to be used in automated detection of disease outbreak and other stress events. This work is the first step in developing a computer-based modeling system for swine feeding behavior.